13
8/4/2019 Articulo CO-CO http://slidepdf.com/reader/full/articulo-co-co 1/13 Idiographic assessment: Conceptual and psychometric foundations of individualized behavioral assessment Stephen N. Haynes a , , Gregory H. Mumma b , Catherine Pinson c a University of Hawaii at Manoa, Honolulu, Hawaii University of Málaga, Málaga, Spain b Texas Tech University, Lubbock, Texas, United States c University of Hawaii at Manoa, Honolulu, Hawaii, United States a b s t r a c t a r t i c l e i n f o Article history: Received 6 August 2008 Received in revised form 20 December 2008 Accepted 22 December 2008 Keywords: Idiographic assessment Clinical assessment Multilevel modeling Assessment instrument development Psychometry Individual differences Idiographic assessment is the measurement of variables and functional relations that have been individually selected, orderived from assessment stimuli or contexts that have been individually tailored, to maximize their relevance for the particular individual. This article rst reviews various de nitions and clinical applications of idiographicassessment.Severalpropertiesof behaviorproblemsand causalrelationsprovidetheconceptualbasis foridiographicassessment: (a) differencesacross persons in theattributes,responsesystems,and dimensionsof a behavior problem, and (b) differences across persons in the causal relations relevant to a particular behavior problem. Because of theseindividualdifferences, nomotheticmeasuresoftenre ectvariance thatis irrelevant to the targeted construct for the individual. We present a psychometric framework for idiographic assessment by rst summarizing why the psychometric principles used to develop standardized measures of nomothetic constructs can create incongruity between the nomothetic measure and the characteristics of the targeted construct for an individual. We then develop a psychometric framework for idiographic assessment that combines components of multilevel modeling (random effects) and con rmatory factor analyses applied to repeated measurements of each individual. We also provide a step-by-step guide for the development and evaluation of an idiographic assessment instrument. © 2008 Elsevier Ltd. All rights reserved. Contents 1. What is an idiographic assessment instrument? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 2. Methods and applications of idiographic assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 3. Conceptual foundations of idiographic assessment I: attributes of behavior problems and their causes . . . . . . . . . . . . . . . . . . . 181 3.1. Individual differences in the attributes of behavior problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 3.2. Individual differences in causal relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 4. Conceptual foundations of idiographic assessment II: psychometric issues in idiographic assessment . . . . . . . . . . . . . . . . . . . . . . . 184 4.1. Aggregate-level optimization and invariant weights of factor and scale scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 4.2. A psychometric framework for idiographic measurement using standardized items . . . . . . . . . . . . . . . . . . . . . . . . . 184 4.2.1. A multilevel random effects factor model: addressing the assessment congruence problem . . . . . . . . . . . . . . . . . 184 4.2.2. Evaluating intra-individual relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 4.2.3. The intra-individual variance/covariance matrix and item-level construct validity. . . . . . . . . . . . . . . . . . . . . . 185 4.2.4. A multivariate, multilevel, random effects, con rmatory factor model for idiographic measurement: overview . . . . . . . . 186 5. Principles and procedures for developing idiographic self-report questionnaires and behavior rating scales . . . . . . . . . . . . . . . . . 188 5.1. Developing the initial array of items for the idiographic assessment template . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 5.2. Procedures for the selection of items for a particular person from the idiographic assessment template . . . . . . . . . . . . . . . 189 5.3. Developing the assessment instrument and assessment strategy for a particular person . . . . . . . . . . . . . . . . . . . . . . . 189 5.4. Initial evaluation of idiographic assessment template, instrument, and strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 5.5. Additional evaluation of psychometric properties of idiographic measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 6. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clinical Psychology Review 29 (2009) 179 191 Corresponding author. Department of Psychology, University of Hawaii at Manoa, Honolulu, HI 96822, United States. Tel.: +1808 956 8108; fax: +1808 956 4700. E-mail address: [email protected] (S.N. Haynes). 0272-7358/$ see front matter © 2008 Elsevier Ltd. All rights reserved. doi: 10.1016/j.cpr.2008.12.003 Contents lists available at ScienceDirect Clinical Psychology Review

Articulo CO-CO

Embed Size (px)

Citation preview

Page 1: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 1/13

Idiographic assessment: Conceptual and psychometric foundations of individualizedbehavioral assessment

Stephen N. Haynes a ,⁎ , Gregory H. Mumma b , Catherine Pinson c

a University of Hawaii at Manoa, Honolulu, Hawaii University of Málaga, Málaga, Spainb Texas Tech University, Lubbock, Texas, United Statesc University of Hawaii at Manoa, Honolulu, Hawaii, United States

a b s t r a c ta r t i c l e i n f o

Article history:Received 6 August 2008Received in revised form 20 December 2008Accepted 22 December 2008

Keywords:Idiographic assessmentClinical assessmentMultilevel modelingAssessment instrument developmentPsychometryIndividual differences

Idiographic assessment is the measurement of variables and functional relations that have been individuallyselected, or derived from assessment stimuli or contexts that have been individually tailored, to maximize theirrelevance for the particular individual. This article rst reviews various de nitions and clinical applications of idiographicassessment. Several propertiesof behaviorproblemsand causalrelationsprovidethe conceptualbasisfor idiographicassessment: (a) differencesacross persons in theattributes,responsesystems,and dimensionsof abehavior problem, and (b) differences across persons in the causal relations relevant to a particular behaviorproblem. Because of these individual differences, nomotheticmeasures often re ectvariance thatis irrelevant tothe targeted construct for the individual. We present a psychometric framework for idiographic assessment by rst summarizing why the psychometric principles used to develop standardized measures of nomotheticconstructs can create incongruity between the nomothetic measure and the characteristics of the targetedconstruct for an individual. We then develop a psychometric framework for idiographic assessment thatcombines components of multilevel modeling (random effects) and con rmatory factor analyses applied torepeated measurements of each individual. We also provide a step-by-step guide for the development andevaluation of an idiographic assessment instrument.

© 2008 Elsevier Ltd. All rights reserved.

Contents

1. What is an idiographic assessment instrument? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1802. Methods and applications of idiographic assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1803. Conceptual foundations of idiographic assessment I: attributes of behavior problems and their causes . . . . . . . . . . . . . . . . . . . 181

3.1. Individual differences in the attributes of behavior problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1813.2. Individual differences in causal relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

4. Conceptual foundations of idiographic assessment II: psychometric issues in idiographic assessment . . . . . . . . . . . . . . . . . . . . . . . 1844.1. Aggregate-level optimization and invariant weights of factor and scale scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1844.2. A psychometric framework for idiographic measurement using standardized items . . . . . . . . . . . . . . . . . . . . . . . . . 184

4.2.1. A multilevel random effects factor model: addressing the assessment congruence problem . . . . . . . . . . . . . . . . . 1844.2.2. Evaluating intra-individual relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1844.2.3. The intra-individual variance/covariance matrix and item-level construct validity. . . . . . . . . . . . . . . . . . . . . . 1854.2.4. A multivariate, multilevel, random effects, con rmatory factor model for idiographic measurement: overview . . . . . . . . 186

5. Principles and procedures for developing idiographic self-report questionnaires and behavior rating scales . . . . . . . . . . . . . . . . . 1885.1. Developing the initial array of items for the idiographic assessment template . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1895.2. Procedures for the selection of items for a particular person from the idiographic assessment template . . . . . . . . . . . . . . . 1895.3. Developing the assessment instrument and assessment strategy for a particular person . . . . . . . . . . . . . . . . . . . . . . . 1895.4. Initial evaluation of idiographic assessment template, instrument, and strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 1905.5. Additional evaluation of psychometric properties of idiographic measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

6. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Clinical Psychology Review 29 (2009) 179 – 191

⁎ Corresponding author. Department of Psychology, University of Hawaii at Manoa, Honolulu, HI 96822, United States. Tel.: +1 808 956 8108; fax: +1 808 956 4700.E-mail address: [email protected] (S.N. Haynes).

0272-7358/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.

doi: 10.1016/j.cpr.2008.12.003

Contents lists available at ScienceDirect

Clinical Psychology Review

Page 2: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 2/13

Idiographic assessment involves psychologicalassessment instru-ments, methods, and strategies designed for an individual respon-dent. Psychologicalassessment canbe individualized in several ways:(a) the assessment strategy (e.g., the assessment instrumentsselected for use and their method of application) can be individua-lized, (b) assessment data can be used to construct an idiographiccase formulation, (c) elements from a nomothetically based assess-ment instrument can be selected to use with a particular respondent,

(e.g., selecting a subset of items from a standardized questionnaire)and (d) elements within a standardized assessment template can beindividualized (e.g., individually selected goals in Goal AttainmentScaling, Kiresuk, Smith, & Cardillo, 1994 , or individually selecteddiscussion topics in an analogue behavioral observation protocol). Inthis article we focus on “c” and “d” above, and discuss the concepts andpsychometric principles that guide the use and development of indi-vidualized assessment instruments.

Idiographic assessment is relevant to a number of psychologicalassessment paradigms but is especially congruent with the behavioraland cognitive-behavioral assessment and treatment paradigm (e.g.,Beck & Emery,1985; Clark,Beck,& Alford,1999; Haynes& O'Brien, 2000 ).Within the latterparadigms, forexample,numerousmanuals have beendeveloped to guide the clinician in using idiographic assessment anddeveloping cognitive-behavioral case formulations (e.g., Beck, 1995;Needleman, 1999; Persons & Tompkins, 2007 ). Despite interest inidiographic assessment, the concepts and psychometric principlesunderlying idiographic assessment have infrequently been articulated.Not surprisingly, then, idiographic assessment instruments haveinfrequently been subjected to rigorous psychometric evaluation.Although we address this paper primarily to the behavioral assessmentparadigm, the concepts, psychometric principles, and methods of development are applicable across assessment paradigms. See, forexample, Barton, Morley, Bloxham, Kitson, and Platts (2005) foridiographic applications of a sentence completion scale.

The goals of this paper are to: (a) review the de nitions, goals, andapplications of idiographic assessment in clinical and research con-texts, (b) provide a precise, parsimonious, useful, and generalizablede nition of idiographic assessment, (c) discuss how the attributes of behavior problems and causal relations are consistent with an idio-graphic approach to assessment, (d) illustrate how a nomotheticassessment instrument introduces error into measures of the targetedconstruct for an individual, (e) describe a psychometric framework forconceptualizing and validating idiographic assessment measures, and(f) recommend procedures for the development and validation of anidiographic assessment instrument. Ultimately, we hope that a moreconceptually and empirically based approach to idiographic assess-ment will promote idiographic science — an increased focus of psychology on the scienti c study of the person (see also commen-taries on the bene ts of an idiographic science of psychology inHoward and Meyers, 1990, and Molenaar, 2004, 2005 ).1

1. What is an idiographic assessment instrument?

Idiographic assessment has been de ned in various ways: (a) Theuse of measures that elicit idiographic information, or content uniqueto the individual case, that can be used to generate case formulations(Barton et al., 2005 ); (b) An individualized method of data collectionand data processing focused on relations between individually-de ned situations and behaviors, designed to give individually-tailored cues for therapy ( Claes, Van Mechelen, & Vertommen,2004 ); (c) Methods, instruments, measures, and contexts designedspeci cally for an individual ( Haynes & O'Brien, 2000 ); (d) Assess-ments designed for the diagnosis and treatment of individual cases, in

which the goal is to make predictions and guide practical decision-making in speci c situations ( McFall, 2005 ), (e) The measurement of constructs or speci c behaviorally de ned targets and pertinentsituational variables that are relevant for a particular individual, andcan serve as indicators of casual relations and/or change over timewithin an individual ( Mumma, 2001 ); and (f) A focus on intraindivi-dual organizationof behaviorin termsof speci c patternsof variabilityin situation-behavior relations over time within an individual ( Shoda,

Mischel, & Wright,1994 ).The de nitions of idiographic assessment have several aspects incommon: (a) at least some elements of the assessment instrument orstrategy are designed to increase their relevance for an individualrespondent, (b) the assessment instrument or strategy is at least par-tially unstandardized , in that some elements can differ across res-pondents, and (c) some elements of the assessment instrument orstrategy (i.e., the template of an idiographic assessment instrument)can be standardized.

Some analogue behavioral observation methods exemplify indivi-dualized assessment within a standardized template ( Heyman & Slep,2004; Snyder, Heyman, & Haynes, 2008; Kern, 1991 ). For example,analogue behavioral observation methods used to measure the verbalinteractions of distressed couples and families often include standar-dized elements, such as instructions to participants, the physicalstructure of the assessment setting, and time-sampling strategies.However, the discussion topics are often based on each couple's orfamily's ratings of the most problematic areas in their relationship toincrease their relevance for that couple or family, (cf. Sheeber, Davis,Leve, Hops, & Tildesley, 2007 , for an example with parent – adolescentdiscussions). For example, Johnson and O'Leary (1996) used the 109item Daily Checklist of Marital Activities (DCMA) as a template foridiographic assessment of behaviors associated with marital satisfac-tion and distress. Each participant selected 20 items from the DCMAand monitored their occurrence daily for seven days.

Our de nition of idiographic assessment emphasizes individualizedmethods of assessment andis congruentwith a rangeofassessment foci,methods, and applications:

“Idiographic assessment is the measurement of variables andfunctional relations that have been individually selected, or derived

from assessment stimuli or contexts that have been individuallytailored, to maximize their relevance for the particular individual. ”2

2. Methods and applications of idiographic assessment

Table 1 illustrates several ways assessment strategies can beindividualized: (a) A subset of items can be selected from astandardized self-report questionnaire or semi-structured interview(e.g., Mumma, 2004 ); (b) In psychophysiological laboratory assess-ment, audio-taped scenarios of trauma-related stimuli can beindividually constructed to match the client's real-life traumaexperiences (e.g. Orr and Kaloupek, 1997 ); (c) The degree to which a

client approximates individually selected treatment goals can bemeasured using a standardized format (e.g., Goal Attainment Scaling,Kiresuk et al., 1994 ); (d) A client can record individually selectedthoughts, actions, emotions, and contexts using electronic diaries(Piasecki, Hufford, Solhan, and Trull, 2007 ); and (e) Individuallyselectedactivities and discussion topics canbe selectedfor couples andfamilies during analogue observation tasks (e.g., Heyman & Slep,2004). As Table 1 illustrates, idiographic assessment is amenable todiverse methods and applications of assessment. Idiographic assess-ment has most frequently been used to measure treatment outcome,but has also been used for case formulation and in research to identify

1 An article by Saul Rosenzweig (1986) in the American Psychologist traces some of the historical roots of an idiographic science in psychology, including Windelband and

Galton's in the 1800s and Gordon Allport in the early and middle parts of the 1900s.

2 Idiographic assessment contrasts with nomothetic assessment, in which judgmentsabout a person are based on comparison with other persons using data from the same

assessment instrument administered in a standardized manner.

180 S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179 –191

Page 3: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 3/13

factors that affect the onset, severity, and duration of behaviorproblems.

A study by Hopko, Bell, Armento, Hunt, and Lejuez (2005) on theeffects of behavior therapy with depressed cancer patients illustratesidiographic assessment in treatment outcome research. Each clientrated his or her weekly progress on 15 individually selected activities(e.g. involvement in social situations, participation in recreationalactivities) on a scale from “easiest to accomplish ” to “most dif cult toaccomplish. ” Weekly evaluations by the therapists of each client'sprogress were used to guide the development of individualized treat-ment programs with weekly behavioral goals.

Because there are important individual differences in the causalvariables and relations for a behavior problem, idiographic assessmentcan also be useful in identifying functional relations (e.g., causalrelations, correlations) between two or more variables relevant to an

individual's behavior problems and in developing a clinical caseformulation for an individual. Haynes, Yoshioka, Kloezeman, and Bello(2009) described a clinical case formulation that was partiallyguided byidiographic assessment, in which the client used an electronic diary toanswer daily questions about individually identi ed marital con icts.

As we address theconceptualand psychometricbases of idiographicassessment in the next sections, we explain how the precision of ameasure (i.e., its accuracy and speci city) can be increased when itselements (e.g. items, scenarios, treatment goals, behavior codes) areselected, developed, or differentially weighted to be more congruentwith the attributesof the targetconstruct for the personbeing assessed.We discuss how idiographic assessment is congruentwiththe nature of behavior problems and causalvariables. Finally, we apply psychometricprinciples and a random-effects analytic model to demonstrate the

potential utility and validity of idiographic assessment.

3. Conceptual foundations of idiographic assessment I: attributesof behavior problems and their causes

In this section we discuss how idiographic assessment strategiesare often appropriate because of: (a) differences across persons in theattributes, response systems, and dimensions of a behavior problem,and (b) differences across persons in the causal relations relevant to abehavior problem.

3.1. Individual differences in the attributes of behavior problems

An idiographic approach to assessment is consistent with hundredsof studiesthat have found signi cant between-person differences in theattributes of behavior problems (see overview of behavior disorders inDiagnostic and Statistical Manualof Mental Disorders, DSM-IV-TR, APA,

2000, and Hersen, Turner, & Beidel, 2007 , and an overview of multivariate models of behavior problems in Haynes, 1992 ). Forexample, personsdiagnosedwith schizophrenia candiffer in the degreeto which they experience impairments in self-care, interpersonal skills,concentration, anhedonia, and affect ( Combs & Mueser, 2007 ).

Because the importance of an attribute of a behavior problem candiffer across persons with that behavior problem, nomothetic andidiographic assessment have different domains of utility. Formaldiagnoses of behavior disorders, such as those based on the DSM-IV-TR, require standardized assessment strategies in order to achieveacceptable levels of between-diagnostician agreement. Nomotheticassessment strategies are also necessary when making population-based inferences, such as in epidemiological research, and to estimateaggregate-referenced probabilities (e.g., which treatment is most likely

to be successful for individuals with a certain disorder). However, as we

Table 1Examples of the methods and applications of idiographic assessment

Reference Participants Measures Assessment method Description Application

Hoff, Ervin,and Friman(2005)

1 pre-adolescentstudent with AHDHand ODD

Disruptive classroombehavior, frequency,antecedences, andconsequences

Behavioral observation (45 min,3– 4 days/week for 6 weeks);Experimental manipulation(alternating baselines with 3treatment conditions)

Hypotheses developed aboutenvironmental factors related todisruptive behavior were testedthrough experimentalmanipulation in the classroom

Alternating treatments design wasemployed to test the impact of manipulating speci c classroomconditions on the frequency of disruptive behavior in order todevelop an individualized

intervention program. Jones-Alexander,Blanchard,and Hickling(2005)

21 child and adolescentmotor vehicle accidentsurvivors, 14 non-accident controls

Heart rate, bloodpressure, skinconductance, subjectiveunits of discomfort

Laboratory psychophysiologicalassessment over 1 stimuluscondition (2.5 – 3 min) and multiplebaselines; Self-report discomfortratings at each condition

Measured physiological andsubjective distress responses toindividuallyconstructed audiotapevignettes of motor vehicleaccidents.

Responses to accident vignetteswere compared for motor vehicleaccident survivors withPTSD, motorvehicle accident survivors withoutPTSD, and non-accident controls.

Lindaueret al. (2006)

24 outpatients withPTSD, 15 police of cerswith PTSD, 15 trauma-exposed non-PTSDcontrols

Heart rate, bloodpressure, severity of subjective anxiety

Laboratory psychophysiologicalassessment, conducted before(2 min), during (2 2 min), and after(4 min) the playing of each of 3scripts; Anxiety self-reportmeasurecompleted after each script

Individually constructed scripts of traumatic, stressful, and neutralevents presented in the lab

Physiological responsivity andsubjective anxiety ratings inreaction to traumatic and stressfulevent scripts were compared foroutpatient and police PTSD groupsand non-PTSD controls.

Mumma(2004)

1 depressed andanxious outpatientadult

Depression and anxietysymptom severity,frequency of relevantcognitions

Self-monitoring questionnaire,completed daily for 90 days

Individualized self-monitoring of anxietyand depressive symptomsand relevant cognitions, using itemsfrom nomothetic measures andidiographic cognitive schema.

Multivariate time-series analysisofdata from dailyratings was usedto determine the psychometricproperties of idiographic cognitiveschema.

Solomon,Arnow, Gotlib,and Wind(2003)

20 females withremitted recurrentdepression, 20 never-depressed femalecontrols

Self-demands regardingself-improvement;number of domains anddegree of self-demand foreach

Self-report questionnaire,administered once

Participants identi ed self-nominated personal shortcomingsfor individually selected self-evaluative domains (out of apossible 16), and rated eachaccording to degree of self-demand.

Scores for remitted depressivesand controls were compared on anidiographic measure of self-demands and a nomotheticmeasure of irrational beliefs.

Wenzel Graff-Dolezal, Macho,and Brendle(2005)

13 socially anxiousuniversity students andtheir romanticpartners. 14nonanxious studentcontrols & romanticpartners

Communication skills,frequency and degree of positive/negativecommunication; socialskills, frequency anddegree of speci c skills

Analogue behavioral observationof 3 different 10-minuteinteractions, conducted for eachcouple

Behavior coding of coupleinteractions during individuallydetermined neutral, problematic,and positive discussion topics.

Socially anxious participants werecompared to nonanxious controlson communication and socialskills.

181S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179–191

Page 4: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 4/13

will discuss, nomothetic assessment strategies are less useful for someother assessment applications, such as evaluating treatment outcomefor a client or a sample of clients, describing the time-course of a client'sbehavior problem, or identifying the functional relations relevant to aclient's behavior problem.

Nomothetic measures often include elements extraneous to theattributes of the targeted construct relevant for a particular person.Because nomothetic measures are selected using item validationprocedures and statistics that optimize relations on the aggregate-

level, when compared to idiographic measures, they tend to introduceahigher proportion of random and nonrandom measurement error forthat individual, therebyreducingtheprecisionof theobtained measures.

Consider a standardized 10-item self-report questionnaire ondepression that includes items on fatigue, early morning awakenings,pessimism, level of physical activity, and depressed mood. Duringdepressive episodes, one client may not experience increased fatiguewhileanother maynot experiencea reduced level of physicalactivity. Anaggregated, standardized depression measure that evenly weights allitems increases measurement error for both clients because variabilityin the scores on the irrelevant items, due to changes in variablesunrelated to depression, contributes to the changes in the “depression ”

score for each client. Consequently, inferences about the clients' level of depression following treatment, or the results of research on the

variables that affect depression, may be compromised. A more precise

and sensitive-to-change measure of depression would be obtained froman individualized questionnaire that omitted the items (or assigned aweight of “0” to the items) that are irrelevant for a client. 3

Fig. 1 illustrates how the degree of measurement error associatedwith a standardized measure for one person is inversely related to thedegree of congruence between the elements of the measure and theattributes of the measured behavior problem for that person. In thecontext of idiographic assessment, “assessment congruence ” is similarto content validity (Haynes, Richard, & Kubany, 1995 ) — the degree to

which elements of the measurement instrument are relevant to andrepresentative of the attributes of the measured construct for aperson. In the context of idiographic assessment, assessment con- gruence is the degree of agreement between the elements of ameasure (i.e., the elements and their weight) and the attributes of thetarget construct for one person. Thus, the degree of congruence for anomothetic measure will vary across persons.

In Fig.1, the behavior problem can include ve attributes, only fourof which are relevant for this client, and only Element C is particularly

3 If an element of an assessment instrument taps an important variable that isirrelevant to the targeted construct but relevant to another important construct, inmany cases we would recommend that this item be retained but not contribute to the

measure of the targeted construct.

Fig.1. A vector diagram illustrating hypotheticalrelations among multiple attributes of a behavior problemand multiple elementsof a standardized measureof thebehavior problemfor one client. Line thickness indicates the relative strength of relation. For this client, the measure of the behavior problem is most strongly affected by “Element D ” but this client'sbehavior problem is most strongly associated with “Attribute C ”. The measure is also affected by Element A, which is not associated with the behavior problem for this client. Thisincongruence reduces the precision of this measure of this behavior problem for this client.

182 S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179 –191

Page 5: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 5/13

relevant (i.e., it accounts for the largest proportion of variance in thebehavior problem). Forthis client, the use of the nomothetic “Measureof Behavior Problem ” as a measure the Behavior Problem includesthree main sources of error. First, variance in the Measure is affectedby variance in “Element A, ” but this element is not associated withvariance in the Target Behavior Problem for this client. For example, astandardized depression questionnaire may include an item on sleepproblems when sleep was not a problem for this depressed client. In

this example, the measure of depression would re

ect variance insleep problems that was not related to depression for this client (butcould be associated with another variable, suchas temporary stressorsat work, alcohol use, or pain from a medical condition).

Second this Measure “Measure of Behavior Problem A ” is moststrongly affected by variance in “Element D ” (e.g., ideas of helplessness)relative tovariance in theotherelements. In contrast, “BehaviorProblemAttribute C ” (e.g., negative automatic thoughts) is the attribute moststrongly related to the target construct for this client. This would beillustrated by 6 items on a questionnairethat queried about “Element D ”

and 3 items measuring “Element C ”. The incongruity between theweighting of elements on the measure and the relative importance of the elements for the individual would reduce the degree to whichvariance in the Measure re ects true variance in the attributes of thetarget construct for this client. 4

Third, Attribute E is associated with the behavior problem for thisclient but variance in this attribute does not contribute to variance inthe measure of that behavior problem. This would be illustrated by aquestionnaire on PTSD that contained no items on nightmares whennightmares were a major problem for the client.

Assessment congruence is estimated by the degree of agreementbetween the weights for the elements of the measure and the weightsfor the attributes of the behavior problem relevant for that person. InFig. 1, there are ve possible agreements — one for each of the threeelements and attributes (B, C, D), shared by both the measure and thetarget construct for that person, one for Element A, which is in themeasure but not a relevant attribute for the person, and one forAttribute E, which is relevant for that person but not included in themeasure. We can quantify the degree of element/attribute agreementby arbitrarily assigning 0 for no agreement or congruence (e.g.,Element A included in measure but Attribute A not relevant for thisperson), a 1 for weak agreement or congruence (e.g., Element Cincluded in the measure but Attribute C is strongly associated withBehavior Problem A for this person), and 2 for strong agreement (e.g.,weighting of Element B and Attribute B are roughly equivalent). Thesum of these element/attribute agreements is 4 (A= 0, B =2, C =1, D =1,E=0) out of a possible 10, so there is a 40% congruence betweenweights of elements of the measure to weights of attributes of thebehaviorproblem.Of course, the degree of assessment congruence willdiffer depending on the particular measure used and the particularindividual assessed.

Notealso that the elementsof a nomothetic measurecan accuratelyre ect the true state of affairs for large samples if optimal instrument

development procedures were followed. However, because congru-ence will usually be less than optimal for any individual in the sample,estimates of the validity of a measure that are derived from largesamples usually overestimate its validity for any individual within thatsample.

An idiographic assessment strategy is designed to reduce measure-ment error and increase precision of the measure by increasing the

degree of congruence between sources of variance in the measure andsources of variance in the target construct. An idiographic measure of “depression ” with greatercongruence for the client in Fig.1 would omitElement A, increasethe impact(weight)of Element C, reducethe impactof Element D, and add Element E.

Idiographic assessment is also consistent with ndings that themost important dimensions (i.e., quanti able attributes of a behaviorproblem such as frequency and duration) of a behavior problem can

vary across persons. For example, episodes of self-injurious behavioroccur frequently and brie y for some children but are less frequentand more prolonged for others ( Mace, Vollmer, Progar, & Mace, 1998 ).To obtain the most precise observational estimate of the behavior, therate of behavior sampling should be directly related to the rate of thebehavior and inversely related to its duration (see discussion of sampling in Minke and Haynes, 2003 ).

In sum, any measure that aggregates across dimensions in thesame way for all persons will have attenuated precision for a par-ticular person because variance in the measure will unduly re ectvariance of unimportant or irrelevant dimensions and insuf cientlyre ect variance in important dimensions. Consequently, the validityand precision of clinical judgments about treatment outcomes, orfunctional relations associated with the measured behavior, will alsobe attenuated.

3.2. Individual differences in causal relations

Idiographic assessment is also consistent with the multivariate andidiographic nature of causal relations for behavior problems. Thou-sands of studies have shown that the same behavior problem can beaffected by different permutations of causal variables and causalrelations, within and across persons. Furthermore,causal relationscanvary over time andas a function of the attribute, dimension, or contextof a behavior problem for an individual client. For example, Hammen(2005) reviewed research indicating that life stressors strongly affectthe likelihood of depression episodes but often have less effect thancognitive variables on the duration of those episodes. Craske andWaters (2005) reviewed research indicating that social settingsstrongly affect the likelihood of a panic attack but often have asmaller in uence than appraisals of physiological sensations on theintensity of the panic attack.

Given the idiographic and conditional nature of causal relations,measurement strategies tailored to the idiosyncratic attributes of aclient's behavior problem can contribute to more precise and clinicallyuseful judgments about causal relations, and thus can facilitate thedevelopment of more effective treatments, a point also made by Johnson and O'Leary (1996) . By contrast, a measure that aggregatesacross behavior problem attributes, dimensions, and contexts in thesame way foreachperson is likelyto provide attenuated andimpreciseestimates of the factors that affect, or are otherwise functionallyrelated to,a particular client'sbehavior problemand thusis morelikelyto lead to less appropriate treatment foci.

Consider an example of differential causal relations for the onsetversus the duration of a client's depressive episodes. If the most im-portant dimension of depressive episodesfor theclient was “frequencyof onset, ” con icts with thepartner would be themoststrongly relatedcausal variable. However, if the most important dimension of depressive episodes for this client was “duration, ” self-criticismwould be the most strongly related causal variable. Consequently,the best treatment foci for this client (e.g., behavioral couplescommunication training versus cognitive therapy) would depend onwhether frequency or duration was the most important dimension of the depressive episodes for this person.

In sum, the precision of our estimates of the causal variables andcausal relationsfor a person'sbehavior problemdependson thedegreeof congruence between the elements of a measure andthe dimensions

and attributes of the behavior problem for that person. Measures that

4 Any error within the elements of a measure, such as items in a scale with lowlevels of internal consistency, could contribute error either on the aggregate level (to astandardized nomothetic measure) or on the intraindividual level (an idiographicmeasure), depending on how item covariance is evaluated. This would be re ectedpartially in the strength of relations between the attributes and the measurement

elements.

183S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179–191

Page 6: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 6/13

fail to differentially focus on the most important dimensions of aperson's behavior problems, or re ect irrelevant dimensions orattributes, will yield imprecise, attenuated, and less clinically usefulestimates of the functional relations for that person.

4. Conceptual foundationsof idiographicassessment II:psychometricissues in idiographic assessment

4.1. Aggregate-level optimization and invariant weights of factor andscale scores

In thissection, wedevelop a psychometric framework for idiographicassessment, starting with an idiographic assessment template contain-ing a set of items covering some or all potentially relevant nomotheticattributes of a behavior problem or constructand using several differentsubsets of these items to assess each person in a sample. We rstsummarize why the psychometric principles used to develop standar-dized measures of nomothetic constructs can create the incongruityproblem illustrated in Fig. 1. We then develop an idiographic psycho-metric framework that combines certain components of a multilevelmodeling approach (random effects) with con rmatory factor analysesapplied to repeated measurements of each individual.

A basic goal of psychometrics is to nd a set of indicators thatmaximize the reliability and item-level construct validity for astandardized measure of a nomothetic construct ( Burns & Haynes,2006; Nunnally & Bernstein, 1994 ). For both the classical and itemresponse approaches, what is “maximized ” is based on relations at theaggregate or group level.

In the classical approach, the correlation of scores between indi-viduals on each item forms the basis for the inter-individual covariancematrix, which is then factor analyzed to determine the dimensionalstructure of the measure and the relative weights of items (the factorloadings). Even thoughthe weights associated with particular items canvary from item to item, each item'sloading on thefactor is essentiallyanaverage “effect ” obtained via estimation methods that are optimal forthe inter-individual covariance matrix and are invariant acrossindividuals. Although this aggregate-level optimization has workedadequately for the purposes of many population-focused psychopathol-ogy and treatment investigations, these nomothetic-level relationshipsarelikely to be sub-optimal foreach individualand, thus, for theaverageof these individuals. As illustrated in Fig. 1, there is likely to be anattenuated congruence between item weights from this generalweightingscheme and the importanceof particularelements,attributes,or dimensions for a particular person.

4.2. A psychometric framework for idiographic measurement using standardized items

We propose a psychometric framework for idiographic measure-ment that accommodates the individualized weighting of elementsnecessary to increase assessment congruence on the level of the

individual and that also is compatible with traditional psychometricsfocused on standardized measures of nomothetic constructs. Referringagainto Fig.1, wedescribea psychometricframeworkthat providestheprecision to assess the different attributes of a construct (e.g.,depression) with maximal congruence for each individual, whiledeveloping aggregate level item weights usable for the psychometricvalidation of a standardized measure of that construct. Thus, thisframework can be used to validate instruments within both theindividualized/idiographic and standardized/nomothetic measure-ment paradigms.

We also describe a validation strategy for this idiographic psy-chometric framework. This validation strategy can be used to guideitem selection for an instrument and evaluate the dimensional(factorial) structure of that instrument on both the idiographic and

nomothetic levels. This validation strategy also assumes that items

intended to measure the target construct are completed by partici-pantson multiple occasions. Thus, thismodel is particularly relevant tovalidation of measures that are to be used for repeated assessments,such as in measuring change within persons across time, response tointervention, or time series/ecological momentary assessment evalua-tion of intra-individual relationships between variables (e.g., whatstressors or thoughts are associated with triggering or maintainingdepressed moodin an individual overtime). Within a researchcontext,

this approach canbe used to validate a measure on both the individualand aggregate levels. Because this approach can be applied to datacollected repeatedly from a single individual, it also can be used toevaluatethe constructvalidityof idiographic measuresdeveloped foraparticular client in a clinical practice setting.

4.2.1. A multilevel random effects factor model: addressing theassessment congruence problem

Thepsychometricframework for idiographicmeasurement thatwepropose uses a multivariate (Mv),multilevel (Ml), random effects (RE),con rmatory factor (CF) model (MvMlRECF). Multilevel modelsinclude a class of models based on multiple regression and a class of models based on con rmatory factor analysis or structural equationmodeling. Our psychometric framework is based on structuralequation modeling (CF analysis). The MvMlRECF model estimates thedimensional structure of a set of items on two (or more) levels: forinstance, the occasion and the individual levels.

For each item, available multilevel con rmatory factor models andsoftware estimate only a single factor loading at each level. Thecovariances among item scores at the occasion level are computed by nding theover-time covariances of the scores among items withinanindividual and then aggregating these covariances across all indivi-duals in the sample. An occasion-level item loading is then estimatedfrom these aggregated covariances. This is a “ xed ” effect in that theover-time dimensional structure of a set of items is assumed to beconstant across all individuals and the incongruity problem shown inFig. 1 still pertains.

Our psychometric framework for idiographic measurementincludes a system for developing and evaluating multi-item measuresthat can be used either in a research or clinical practice context. Toavoid the congruence problems illustrated in Fig. 1, the weighting(factor loadings) of items must be individualized for each person andthus regarded as “random ” effects. However, we know of no way of obtaining empirical estimates of random-effect factor loadings usingscores obtained from a single occasion. The individualized factorloadings for each of the items measuring each construct must beestimated from intra-individual scores obtained over multiple occa-sions within each individual ( Molenaar, 2004 ). Such scores can beobtained with two basic kinds of designs — experimental and passive-observational (see Morgan and Winship, 2007 , for additional discus-sion of experimental and passive-observational designs).

4.2.2. Evaluating intra-individual relationships

The “gold standard ” for evaluating intra-individual relationshipsinvolves interrupted time-series designs in which hypothesized causalvariables are systematically manipulated while monitoring targetbehaviors. Shadish, Cook, and Campbell (2001) and Hayes, Barlow, andNelson-Gray (1999) discussed several types of interrupted time-seriesdesigns, such as reversal (e.g., ABAB), multiple baseline, and multi-element designs. For example, the additive effects of prosocial role-playing and contingent reinforcement for prosocial play of an aggressivechild couldbeexaminedin anA 1B1A2B2 design, with A 1 =2 weekbaselineno intervention, B 1 =4 week role-play and reinforcement intervention,A2 =1 week return to baseline conditions, andB 2 =reinstatement of role-play and reinforcement intervention. Multiple daily measures of severaloutcome variables for each subject could include: (a) observations byexternalobserversof the rate of approaches to other children, therate of

approaches to target child by other children, and duration of prosocial

184 S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179 –191

Page 7: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 7/13

play; (b) ratings by teachers of the child's prosocial behavior initiationsand skills; and (c) reports by the target child of frequency of, andsatisfaction with, social contacts with peers.

In theprocess of evaluatingsuchintra-individualfunctionalrelationswith interrupted time-series designs, multiple measures of a targetvariable areobtained repeatedly foreach individual. These dataare usedto obtain covariances among the measures so that intra-individual itemloadings of the measures on the latent variable of interest can be

obtained. In so doing, items with optimal congruence and sensitivity foreach person can be identi ed. In the example just described, theoutcome variable “prosocial play ” was operationalized with severalobservational and self-report measures. If an aggregate, multi-itemoutcome measure of prosocialplaywere to beconstructed,an analysisof the individual covariance structure would indicate which of thosemeasures provided the best estimates of the “prosocial play ” constructfor each child (i.e., which measures evidenced the highest congruenceand sensitivity to change with the targeted construct).

In contrast to this type of interrupted design, many types of functional relationships of interest to cognitive-behavioral clinicianscannot be conveniently evaluated with interrupted time-seriesdesigns. Cognitive, life-stressor, and other putative causal variablescan bedif cult to manipulate and effects areoften delayed,prolonged,and not easily reversed. Yet we still seek to evaluate functionalrelations among life-stressors, cognitions, affect, behavioral responsesequences, and dysfunctional behavior using measures that aremaximally sensitive and congruent for a particular client. In suchsituations, the validation of a multi-item idiographic measure usesmultivariate time-series regression designs , in which multiple variablesare monitored frequently across time, without any manipulation, toestablish time-lagged and concurrent functional relations (see, forexample, a multivariate time-series regression study by Thatcher andHaynes, 2001 , in which multiple measures were obtained across60 days from rheumatoid arthritis patients).

4.2.3. The intra-individual variance/covariance matrix and item-levelconstruct validity

Whether from interrupted time-series designs or multivariatetime-series regression designs, scores fromthe repeatedmeasurementof a variable within an individual, using multiple measures of thatvariable, canbe used to estimate theextentto whichscores covaryovertime intra-individually. For example, for Person 1 in Table 2 , “Cried alot ” would covary over time strongly with “Felt down, blue, ” but onlymoderately with “Not interested in things, ” and not at all with“Appetite change. ” The over-time covariances for the setof items forman intra-individual variance/covariance matrix, which shows theinterrelations of the item scores over time within the individual(Hershberger, 1998; Molenaar, 2004 ). Factor analysis of the intra-individual variance/covariance matrix can yield factor loadings foritems that are maximally sensitive (congruent) indicators of the targetconstruct for that individual. These factor loadings are used to evaluateitem-level construct validity ( Burns & Haynes, 2006; Mumma, 2004;

Mumma & Mooney, 2007a ).In its simplest form, the intra-individual variance/covariance

matrix shows concurrent (same time) relations among multiplemeasures of the same variable. Within each person, convergent

validity is demonstrated when multiple measures of a construct arehighly intercorrelated and strongly load on that factor. Item-leveldiscriminant validity is demonstrated when the measures do notsigni cantly load on other factors. Table 2 illustrates intra-individualitem communalities (i.e., squared loadings) for four items on a“depression ” factor for four items. For Person 1, “Cried a lot ” and “Feltdown, blue ” have relatively large loadings, indicating that thesefactors are good indicators of depression for this person. “Not

interested in things”

has a moderate loading indicating that it is anacceptable convergent measure of depression for that person.However, “Appetite change ” has a psychometrically “non-signi cant ”

loading (below the typically used loading cutpoint of .40) indicatingthat it is not a good indicator of depression for this person. Person 2has a different pattern of loadings — “Cried a lot ” and “Appetitechange ” are highly congruent (high loadings). However, “Felt down,blue ” has low item-level convergent validity for Person 2 and thus isnot a valid intra-individual indicator of depression for this person.Therefore, the loading for each item varies from person to person,depending on the congruence of that item to the most importantattributes of the target construct for each individual. Due to thevariability between persons, these loadings constitute a randomeffect. Multilevel model software can estimate the variance of thisrandom effect and test whether or not it is statistically signi cant.

There will be an intra-individual variance/covariance matrix foreach individual in the sample. For each intra-individual variance/covariance matrix, factor loadings are estimated using an intra-individual con rmatory factor analysis such as P-technique factoranalysis. Provided there is variability in the target behavior over timeor across contexts, these loadings provide a measurement pro le (i.e.,item weights) that is maximally congruent with the most importantattributes of the behavior problem (construct) for an individual duringthe interval when the measures were taken. Note that more complexstructures for the intra-individual variance/covariance matrix andthus for the intra-individual factor analyses are possible. For example,the intra-individual variance/covariance matrix could include auto-covariances and cross-lagged covariances between items at one ormore lags (for examples, see Hershberger, 1998; Molenaar, 1985;Wood & Brown, 1994 ). The con rmatory dynamic factor analysisconducted on the intra-individual variance/covariance matrix for eachperson would be sensitive to dynamic effects between factors (latentvariables) or between item scores and lagged factors over time withineach person, depending on the speci c con rmatory dynamic factormodel used ( Nesselroade, McArdle, Aggen, & Meyers, 2002 ; seeMumma, 2004; Mumma & Mooney, 2007a for examples of con- rmatory dynamic factor analysis applied to multivariate time seriesdata from single individuals).

Either approach – P-technique factor analysis or con rmatorydynamic factor analysis – produces item loadings speci c to eachindividual. Both models are con rmatory because each item isconstrained to load only on itstargeted factor, although themagnitude(or even direction) of the loading will vary across individuals.

This intra-individual variance/covariance matrix validation strat-egy can be applied in either a research or clinical context. In a clinicalcontext, these analyses can be used to evaluate construct validity(item level convergent and discriminant validity) and determine theitem loadings to produce idiographic measures with optimalcongruence for a particular individual. This has been demonstratedusing both formal con rmatory dynamic factor analysis ( Mumma,2004; Mumma & Mooney, 2007a,b ) as well as using shorter time-series and simpli ed statistical models and software that would bereadily accessible and useable by clinicians ( Smith & Mumma, 2006;Mumma & Fluck, 2009 ). In either case, once a congruent measure of the target construct for that individual is obtained, relations betweendifferent variables or constructs (e.g., cognitions, stressors or situa-tional triggers, distress) relevant for that individual can then be

evaluated more precisely.

Table 2Hypothetical item communalities (squared loadings) of four items on a depressionfactor for three individuals, obtained from daily measures across 60 days

Person

1 2 3

ItemCried a lot .68 .82 .12Felt down, blue .80 .14 .44Not interested in things .32 .41 .54Appetite change .09 .65 .47

185S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179–191

Page 8: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 8/13

In a research context, this strategy to develop idiographicmeasures can be used to develop optimal (i.e., congruent) itemloadings/weights for each person in a sample. The second level of themultilevel model then produces estimates of the loadings for eachitem, such as the four items in Table 2 , averaged across individuals inthe sample (e.g., in Table 2 , the average squared loading for “Cried alot ” =.54). Alternatively, within each individual, the score on eachitem can be averaged across occasions (time), and this average score

for each person is used in the level 2 analysis, to obtain an averagelevel of depression across individuals in the sample. As with anyevaluation of construct validity, the validity of this approach can betested by showing the expected convergent or discriminant relation-ships of the target measure to other measures across time ( Furr &Bacharach, 2008 ). For example,we could examine the degreeto whichscores on this individualized measure of depression correlate in theexpected direction and magnitude – over occasions within theindividual – with other measures of distress (e.g., a measure of anxiety) or with cognitions hypothesized to trigger or maintaindepression for some or many individuals in the sample. Since theserelations are obtained within each individual, we can obtain both theaverage strength of these intra-individual relations and examine thedegree to which these relations differacross persons (a level 2 randomeffect in the multilevel model) and across occasions (i.e., we couldexamine homogeneity of covariance across time).

We can also compare this idiographic measurement approach to astandardized nomothetic approach in which the items are unitweighted (or otherwise equally weighted) for all persons. Comparedto a nomothetic measurement strategy, functional relations (e.g.,between speci c events and “depression ”) should be more accuratelyidenti ed with the idiographic measurement approach. However,future research must show that the idiographic measurementapproach is incrementally valid compared to a nomothetic measure-ment strategy at both the overall test score and facet levels ( Haynes &Lench, 2003; Smith, Fischer, & Fister, 2003 ).

It is important to note that, except under a rather restrictive set of conditions, the relations between item scores measured over time(occasions) within individuals will almost always be different fromtherelations between items scores using aggregate-level cross-sectionaldata ( Molenaar, 2004; Molenaar & Valsiner, 2005 ). The intra-individual variance/covariance matrix will differ from the inter-individual covariance matrix except under conditions that are quiterestrictive (see discussion in Molenaar, 2004 ). Thus, the results of thetypical R-technique factor analysis applied to the inter-individualcovariance matrix – whether con rmatory or exploratory – will likelybe quite different from the con rmatory dynamic factor analysisperformed on the intra-individual variance/covariance matrix of eachindividual. We will now describe the ML or random effects modelwhich serves as the basis for the psychometric theory of idiographicmeasurement.

4.2.4. A multivariate, multilevel, random effects, con rmatory factor

model for idiographic measurement: overviewUsing the intra-individual variance/covariance matrix from each

individual, the level 1 model estimates the loading of each item on itstarget construct for that person's scores on items collected overmultiple occasions. The model is developed for J constructs ( j=1 to J )and k items within each of these constructs. Item scores are collectedover time for each individual. As described above, the loading of anitem on itstarget constructis individualizedfor each person, a randomeffect across individuals.

This model assumes that we are measuring the same constructs(factors) and facets thereof, for each person and that each targetconstruct is unidimensional. For example, using this idiographicmeasurement strategy, each person rates the same set of items fordepression and the same set of items for anxiety, with the two sets

containing different items. The number of items within each set and

the number of occasions on which scores are obtained need not beequal. 5 The constructs or factors can either be related or independent.

4.2.4.1. Level 1 model. In this MvMlRECF model, the rst level isoccasions and the second level is individuals. The level 1 model is:

Y ijkt = A0ijk + λ ijk X ijt + nijkt ð1Þ

where

Y ijkt is the score for person i on construct j for item k on occasion(e.g., day) t , j=1 to J , k=1 to K j..

A0 ijk is the intercept for the ratings for person i on construct j foritem k when the latent variable or factor score X ij forconstruct j is at it's mean for that person. This is theexpected rating on item k (e.g., “felt sad ”) given the averagerating of the latent target construct j (e.g., depression) forperson i.

λ ijk is the loading of item k on construct j for person i. X ijt is the “score ” on latent variable (construct) j for person i at

time t .n ijkt is the random shock, noise, or residual for person i, on item

k of construct j at occasion t .

Each target construct (e.g., depression, anxiety) is conceptualizedas a commonfactor or latent variable X j. Each is measured by the over-time covariance shared by the items potentially targeting thatmeasure. A unique and essential feature of this MvMlRECF idiographicmeasurement framework is that the importance of particular itemsvariesacross people so as to maximize congruence foreach individual,within constraints described later. Technically, because the loadingson a given item differ across individuals, the level 1 residual variancesdiffer across items and individuals ( σ e|| ijk

2 ).Several additional features of this model warrant discussion. First,

there is no need to collect data on all items within each construct forall participants. If a particular item is not included for certain indi-viduals, the λ

ijkis simply not estimated for that item for that person.

This unestimated loading is then regarded as missing data, which doesnot present a problem for this model if the pattern of missings isassumed to be missing at random ( Little & Rubin,1987 ). Alternatively,for a particular individual, if an item is rated but has insuf cientvariance (perhaps due to a oor or ceiling effect), the loading for thatitem can be set to 0, so that its contribution to the measurement of thetarget construct for that person is included in the level 2 model.

Second, the occasions when the data are collected need not be thesame across individuals. Occasions are considered nested within eachindividual and the data could be collected during March for oneparticipant andMay foranother. However, forthe level 2 analysis to bemeaningful, the same sampling plan or frequency should be used forall individuals (e.g., daily ratings; xed-interval time sampled EMAratings). Otherwise, the interval for, and meaning of, a lag in thecon rmatory dynamic factor analysis is unclear. Importantly, though,once this general over-occasion(over-time) sampling plan is speci ed,the speci cs of data collection/sampling can be individualized for eachperson. For example, one person could complete daily ratings atdinnertime whereas another person could complete them at bedtime.Morecomplicated occasionsampling plans, suchas randomlysignaledtimes or event-sampling methods, are potentially tractable butbeyond the scope of this paper.

5 The minimum number of occasions is related, in part, to the number of items foreach target construct, the number of lags represented in the intra-individualcovariance matrix, the number of lags over which dynamic relationships in thecon rmatory dynamic factor model are to be estimated, and the degree of variability in

scores on each item for that person.

186 S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179 –191

Page 9: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 9/13

A nal feature of this level one model is that each latent variable Xjcan be conceptualized as a factor with a dynamic structure. Thus, the(estimated) factor score at time t may be related to the factor score onthe previous occasion ( t −1). In a con rmatory dynamic factor analysismodel, this dynamic relationship can be modeled as different itemsloading differently at different lags (the White Noise Factor Scoremodel) or as the latent variable having the same concurrent factorloading structure across lags, while the dynamic structure is captured

with an autoregressive parameter across lags of the latent variable(the Direct Autoregressive Factor Score model: Nesselroade et al.,2002; Mumma, 2004 ). Using the latter model – which is generallyconceptually and statistically simpler and more parsimonious – thelevel 1 model for a CDFA with one lag would become:

Y ijkt = β 0ijk + λ ijk X ijt + nijkt and

X ijt = β j X ij t −1ð Þ+ dijt

ð2Þ

where β j is the autoregressive parameter estimating the extent towhich scores on thelatent variable(factor) from theprevious lag( t −1)predict scores on the latent variable now ( t ), for each person j. A directautoregressive factor score model on two constructs for concurrentand lag 1 time frames would involve estimating two autoregressiveparameters, one for each factor, and perhaps two cross-laggedcovariances between the two latent variables (e.g., depression andanxiety) for each person.

4.2.4.2. Level 2 model. An importantfeature of theMxMLRECF modelis that the factor loadings for each item λ ijk vary randomly acrossindividuals. This random effects component of the model is capturedin the level 2 model:

A0ijk = γ 00 + u0ijk ; and

λ ijk = λ jk + u1ijk ;

ð3Þ

where

γ 00 is essentially the grand mean of the intercepts for item jk,collapsed across individuals who have occasion-level datafor that item,

u0 ijk is the level 2 random effect associated with A0 ikj , theserepresent the different item means treated as random effectacross items and individuals,

λ jk is the average loading of item j on construct k, essentially aweighted averageof the λ ijk across the individuals who haveoccasion-level data for that item, and

u1ijk is thelevel 2 random effect associated with λ ijk . Forexample,for item 1 of target construct 1 for person 1, u1111 is how farthe intra-individual factor loading for this person on thatitem ( λ 111 ) is from the average loading λ 11 on that target

construct ( X 1).

A relatively low average loading λ jk for one item suggests that,across persons in this sample, the item is a less important indicator of thetarget constructthan theotheritems used. A small averageloadingon many or all of theitemssuggeststhat, across persons in the sample,the setof items is not appropriately sampling the facets or elements inthe domain of the target construct. Alternatively, the target constructmay not be suf ciently coherent, suggesting a need for greaterspeci cation. Further details about the level 2 residuals u1 ijk areavailable from the second author.

4.2.4.3. Limitations: psychometric and statistical. Several psycho-metric and statistical limitations of this psychometric framework for

idiographic measurement should be noted.First, at this time, we know

of no software that can directly estimate this MvMlRECF model. Thus,estimation would need to occur in two separate stages. First, aseparatecon rmatory P-technique or dynamic factor analysisneeds tobe run on the sets of items for the target constructs foreach individualto obtain the person-speci c loadings ( λ ijk) for each item on eachconstruct. This could be readily done as a multigroup analysis usingwidely available structural equation modeling software where the“group ” is the individual and the data is the intra-individual variance/

covariance matrix for each individual. Second, these loadings wouldthen need to be entered into a separate program that would estimatethe variance of the loadings for each item ( σ u 1 jk

2 ) and the covariancesbetween the random effect for each item. With more than just a fewitems for each construct, constraints will probably need to be placedon these covariances. 6

A second issue is the time frame or interval over which aninstrument is validated using the psychometric framework andvalidation strategy described above. Daily ratings, for instance, canbe examined for day-to-day variability and item covariance, weekly ormonthly cyclicity, or longer-term linear or quadratic trend. The intra-individual variance/covariance matrix is formed based on covariancesthat may confound one or more of these sources of variance (e.g.,weekly cyclicity and daily variability of depression scores may both bepresent in the variance/covariance matrix). However, a number of statistical strategies are available to remove such confounding —

including removing longer-interval covariance (e.g., trend, monthlycyclicity) via statistical partialling and removing shorter-intervalcovariance via smoothing (for examples, see, Hokanson, Tate, Niu,Stader, & Flynn, 1994; Mumma, 2004 ). Essentially, the investigator orclinician can choose to focus on item covariances occurring at thefrequency that is of greatest theoretical or practical interest, such asdaily, weekly, monthly, etc.

Thirdly, the strategy outlined above presumes that items areappropriately constructed and are at least potentially adequatemeasures of the target constructs for at least some individuals. Itemconstruction issues such as item ”dif culty ” or severity level (e.g.,wording the item as ”feel like crying occasionally ” versus ”haveuncontrollable lengthy spells of sobbing ”) and response scale (e.g.,number of rating options, appropriateness of verbal anchors) need tobe addressed so as to increase sensitivity to the severity range of individuals represented in the target population. Also, given that,ideally, a standardized set of items is given to each person in thisapproach, the items need to be selected to capture the most relevantfacets, elements, or attributes of the target construct (e.g., dysphoricmood, loss of interest, or anergia for depression) for as many of theindividuals as possible. As mentioned, a generally (on average acrossthe sample) poor item can be detected within the MvMlRECFframework by an average loading λ jk that is relatively low, althoughsuch a result is conditional on what other items are used to measurethe target construct.

A nal psychometric and statistical limitation is that implementingthis random effects psychometric model for idiographic measurement

as a two-stage estimation procedure may not provide empirical bayesestimates of the random effect loadings for the items for eachindividual. Thus, the estimate of how far the loading for a particularitem for a particular person is from the mean loading λ jk is notregressed to themean as a function of thereliability of theestimate forthat person.

4.2.4.4. Limitations: practical issues. Practical limitations includeclinician and participant time and response burden. Advances inautomated (web-based) and ambulatory (PDA) data collection havegreatly decreased some of this burden. Also, the automatic storage of

6 Free estimation for, say, 8 items (4 for each of 2 constructs) would result in 28covariances for the concurrent relationships alone (plus 56 cross-lagged covariances

and 8 autocovariances if a lag 1 intra-individual variance/covariance matrix is used).

187S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179–191

Page 10: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 10/13

data available with many data collection instruments greatly reducesthe burden of data entry ( Shiffman, 2007 ).

4.2.4.5. Summary. We have presented the conceptual and statisticalfoundations for a random effects psychometric model for idiographicmeasurement. At the present time (due largely to limitations inavailable software), implementing this model is a rather complex andtime-consuming process. However, with advances in structuralequation and multilevel modeling software, these limitations maybe ameliorated in the future. Nevertheless, we believe this model isthe rst one to adequately describe a psychometrics of idiographicmeasurement that resolves the congruity issue illustrated in Fig. 1.

5. Principles and procedures for developing idiographicself-report questionnaires and behavior rating scales

In this section we describe how to initially develop an idiographicassessment instrument, based on the conceptual and psychometricprinciples presented in the previous sections. First, we describe howto construct an “idiographic assessment template ” that can be used todevelop individually tailored self-report questionnaires or ratingscales. Although we focus on self-report questionnaires and ratingscales, the methodology outlined below is also applicable toidiographic assessment instruments developed within other assess-ment methods, such as behavioral observation, self-monitoring,laboratory-based psychophysiology, or ambulatory biomeasurement.

The idiographic assessment template is a standardized assessmentinstrument used to select the most relevant elements of the targetvariable (construct) for a particular person. Fig. 2 illustrates a smallsection of a larger idiographic assessment template, the Prior Intimate

Partner Stalking Scale , used in a time-series research program on the

effects of stalking. The full template includes 120 potential stalkingevents drawn from the stalking literature. Because a stalker typicallyengages in only a few of these behaviors, each subject (a currentvictim of stalking) in the study monitors a small sample of theseevents daily. Each item assesses an element and attribute of stalkingrelevant to that person and is individually selected from the templateon the basis of its frequency of occurrence as reported by the subject. 7

Idiographic assessment templates are most effective when theyfocus on a homogeneous behavior problem, causal variable, orpopulation. The idiographic assessment template should include anarray of possible assessment targets, only some of which will berelevant for each respondent. If the items in the template are notcon ned to a homogeneous construct (or facet thereof), error will beintroduced into the idiographic measure.

The sections below describe the process of developing the initialarray of items for the template, procedures for the selection of the

methods for evaluating the content validity of the idiographicassessment template, the process of developing an individualizedmeasure and assessment strategy for a particular person, andevaluation of the construct validity of the scores for each person. Insome cases, a standardized nomothetic assessment instrument mayhave suf cient content validity (the elements of the instrument coverthe facets of the targeted construct and none of the elements coverextraneous constructs) and precision to serve as a template. Thestrategies outlined below are re nements, for applications in anidiographic assessment context, of widely accepted procedures for the

Fig. 2. An abbreviated example of an idiographic assessment template. Each respondent will monitor daily a different set of stalking events, based on her report of the frequency of currently occurring stalking events on the template.

7 The Prior Intimate Partner Stalking Scale was developed by Karen Kloezeman, at theUniversity of Hawaii, Manoa. In this study each participant will monitor the occurrenceand severity of stalking events, along with mood, anxiety, sleep, and intrusive

thoughts, daily for 60 – 90 days.

188 S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179 –191

Page 11: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 11/13

development and content validation of nomothetic assessmentinstruments (e.g., Haynes et al., 1995 ).

5.1. Developing the initial array of items for the idiographic assessment template

When there is no single assessment instrument with satisfactorycontent validity that can be used as a template, the rst task in the

development of an idiographic assessment instrument is to develop acomprehensive array of items (e.g., potential questionnaire items,potentialitems fora ratingscale) forthe template.The items shouldberepresentative of the targeted behavior problem, population, orassessment focus, and all items should be relevant to the targetedconstruct. Different subsets of this array will subsequently be selectedfrom the template for each respondent. Template developmentrequires careful speci cation of the assessment targets and intendedapplications of the idiographic assessment instrument.

1. Specify the target construct and application of the idiographicassessment instrument: Specify the targeted behavior problem, population, events, and goals of the idiographic assessment instru-ment. Examples of these categories include treatment outcomemeasurement of children with ADHD or adults with social anxiety,

daily monitoring of positive and negative life events, or variablesassociated with the onset of panic episodes.

2. De ne the domain , attributes, and dimensions of the targetedconstruct to be included and excluded. For example, attributes of panic attacks to be assessed may include catastrophic thoughts,subjective experiences of distress, psychophysiological responses,and protective behaviors. In addition, dimensions of interest forpanic attacks of interest may include severity, frequency, orduration of symptoms. Domain refers to the limits of the targetedconstructs that are to be measured (e.g., will the instrument alsoinclude items on contexts and settings for panic episodes, or becon ned to associated symptoms?).

3. Establish the array of potential items for the template. Examplesinclude an array of potential symptoms of depression or an array of

potential verbal communication problemsfor a couple. The array of potential items should be congruent with the domains, attributes,and dimensions speci ed for the targeted construct. The array of potential items for the template can be derived from existinginstruments in combination with additional item generationprocedures, such as suggestions from experts and persons fromthe targeted population (see Haynes et al., 1995 , for extendeddiscussion of item generation procedures).

4. Ensure that the items in the array of potential items are relevant andrepresentative for the targets, functions, applications of the assess-ment, and for the domain, attributes, and dimensionsof the targetedconstruct (i.e., the items have good content validity ). Ensure that thearray of items includes all potentially relevant variablesand excludesirrelevant variables. Expert ratings of the relevance of items andrepresentativeness of the array can assist in this process.

5.2. Procedures for the selection of items for a particular person from theidiographic assessment template

Once a relevant and representative array of items has been gene-rated, following procedures 1 – 4, procedures for the selection of themost relevant items for each respondent should be developed. Theselected items will provide the content of the idiographic assessmentinstrument for a particular respondent.

5. Specify the response format and scale (e.g., 4-pt Likert scale orordinal categories) that will be used in the idiographic assessmenttemplate. The response format and scale on the idiographicassessment template may depend on whether the severity,

duration, or the rate of occurrence is the most relevant dimension.

In Fig. 2, the response format includes 4 categories that cover theexpected range of stalking behavior rates for this population,selected on the basis of prior research. The response format andscale should be designedto maximizesensitivity for the individualsin the target population.

6. Develop appropriate instructions for item ratings on the idio-graphic assessment template. For example, “Read each itemcarefully and circle the appropriate response for how often you

experienced each behavior”

(see Fig. 2).7. Establish criteria for the selection of items from the idiographicassessment template array thatare to be included in the idiographicassessment instrument foreach person.For example, items from thearraycouldbe selected on thebasisof a severity or frequencycriteria.For the template illustrated in Fig. 2, the ve most frequentlyoccurring stalking events, out of 120 possible, are selected formonitoring by each person. Selection criteria should re ectconsiderations such as content validity for each client (coveringmajor attributes of the targeted construct for each client), sensitivityto change and the assessment load on the client. Note that thenumber of variables selected can differ across clients but the mainpurpose is to include only those that are most relevant and excludethose that are least relevant for each person.

5.3. Developing the assessment instrument and assessment strategy for a particular person

Procedures 1 – 7 result in an idiographic assessment template thatincludes an array of potential items, a response format and scale,instructions to respondents, and criteria for selecting the most relevantitemsforeach person. Theclinician or researcher must also establishthespeci c assessment strategy for idiographic assessment. The strategycan include the timing of assessments, the occasions or contexts of assessment, how data are collected, and methods of monitoring ad-herence in the case of repeated assessments. Some elements of thestrategy can be individualized while others are standardized. Forexample, for the stalking assessment each respondent monitors a

differentset of stalkingevents, butall respondents recordthe eventsandother variables each evening before going to bed. Similarly, individua-lized treatment outcome questionnaires would all be administered atthe termination of treatment or at speci ed follow-up times.

Strategies of assessment will also depend on thespeci c method of data acquisition and the instruments used. For example, strategies foracquiring daily data on mood would differ depending on whetherpaper-and-pencil or hand-held computers were used.

8. Specify the times and/or settings in which assessment is to occur(e.g.,at home,in theclassroom,aftermeals,randomly throughouttheday). Thesettings should be selected based on their relevanceto the individual as well as to the targets and goals of assessment.Forexample,behaviorproblems,such as those related to marital/family interactions, may be more likely to occur at home in the

evenings. Alternatively, the assessor may want to comparebehavior across different settings. The idiographic assessmentinstrument should have a section where settings, times, and/orcontexts can be speci ed for the respondent.

9. Select an appropriate format for the instrument (e.g., font sizeand type, spacing, and visual arrangement of the instrument).Usually, this will be standardized across respondents.

10. Develop instructions for respondents that are appropriate forthe items, methods, strategy, and functions of the assessment.The instructions, and other elements of the template, shouldalso be appropriate for the developmental level, cognitivecapacity, and special needs of the respondents.

11. Establish methods to monitor adherence by respondents to theassessment strategies. This will depend on the method of

assessment but might include daily mailing of data forms,

189S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179–191

Page 12: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 12/13

phone delivered data, weekly contacts, or time stamps onpalm-top computers. The method of monitoring adherenceshould be appropriate to the person and might depend onconsiderations such as an individual's ability to understand andaccept the assessment protocol, dif culty in remembering tocomplete repeated assessments, schedule/workload or othercompeting demands, and other pragmatic issues or encum-brances that may interfere with an individual's adherence.

5.4. Initial evaluation of idiographic assessment template, instrument,and strategy

Procedures 1 – 11 are designed to produce a parsimonious idio-graphic assessment template (procedure 1 – 4), instrument (procedure5– 7), and strategy (procedure 8 – 11) with high levels of contentvalidity. Procedures 12 – 13 are designed to evaluate the contentvalidity and practicality of the idiographic instrument.

12. Have experts provide quantitative and qualitative evaluation of allaspects of the idiographic assessment template and instrumentdevelopment (1 – 11above). For example, experts could be asked,“Please rate the degree to which each item in this template arerelevant to ( … ) on a 4-pt scale (Very Relevant, etc). and, “Are thereadditional items that should be included in this template? ”

Additional queries would focus on the instructions, selectioncriteria, response format, and other elements of the assessmentstrategy.

13. Re ne the elements of the idiographic assessment template,instrument, and strategy on the basis of feedback from experts,and re-evaluate in iterative fashion. We also recommend pilot-testing prior to formal implementation in research applications.

5.5. Additional evaluation of psychometric properties of idiographic measures

Once the data are collected from the idiographic instrument,additional psychometric properties of the obtained measures can beexamined with the MvMlRECF model presented in the previoussection. If the primary focus is on one individual in a clinicalassessment context, procedure 12 and 13 may be suf cient. If theprimary focus is on oneor more individuals as part of a research study,then procedure 14, described below, would also be used to evaluatethe item- and construct-level convergent and discriminant validity of the target construct(s).

14. Evaluate intra-individual construct validity of items completed byeach individual. These analyses can be done more ef ciently if data was collected and automatically stored using a web- or PDAbased platform. For each item, scores over time should be brie yexamined in a graph (e.g., in Excel) so that items with inadequatevariability (e.g., oor or ceiling effects) are removed. Item-level

convergent validity is evaluated using correlations, ideally bydoing a con rmatory P-technique factor analysis (for concurrentrelationships) or con rmatory dynamic factor analysis (forconcurrent and lagged relationships, see e.g., Mumma, 2004 forspeci cs). Items with low loadings (e.g., b .40) might be deletedwhen creating the scale score for that individual, although suchloadings should be retained if a level 2 (across individuals)MvMlRECF model is being estimated for a multi-person study.

Relations between subscales (e.g., measuring different facets of thetarget construct) are also evaluated using correlational methods asdescribed above. For example, items within an attribute should bemorestrongly correlated thanitems betweenattributes so that,ideally,each item should load on only the targetedattribute. Finally, predicted

functional relations between the target construct and other variables

are examined, all using intra-individual data. A latent variableapproach inherently gives a higher weight to items that are the bestindicators of the construct for that individual, thereby increasing thecongruence between the measure's elements and the behaviorattributes for that individual (see Mumma & Mooney, 2007a,b , for anillustration of this method).

15. If the numberof itemsused for theidiographic assessment of each

participant is relatively large compared to the item pool (e.g., eachperson completes at least 10 of 20 potentially useable items), thelevel 2 MvMlRECF model should be estimated. The xed effect istheaverageloading of a particular item and therandom effects arehow variable the loading of each item is in measuring the targetconstruct across individuals in the sample. If this level 2 residualvariance is statistically signi cant for an item, the individualweights rather than the uniform (aggregated) weight should beused in subsequent analyses of study hypotheses.

6. Summary

We reviewed the de nitions, goals, and applications of idiographicassessment in clinical and research contexts, de ned idiographicassessment in a way to maximize its utility, applicability, andprecision, discussed the conceptual foundations of idiographicassessment, explained how a nomothetic assessment instrumentintroduces error when applied at the level of the individual, describeda psychometric framework for conceptualizing and validating idio-graphic assessment measures, and outlined several steps in thedevelopment and validation of an idiographic assessment instrument.

Idiographic assessment is the measurement of variables andfunctional relations that have been individually selected, or derivedfrom assessment stimuli or contexts that have been individuallytailored, to maximize their relevance for the particular individual.Underlying the use of idiographic assessment is the concept of as-sessment congruence — the degree of agreement between the elementsof a measure, including their weight, and the attributes of the targetconstruct for the person being measured. Because there are importantdifferences across persons in the attributes of a particular behaviorproblem, and in the causal variables and relations relevant to thatbehavior problem for that person, the degree of congruence for anomothetic measure will vary across persons but will almost alwaysbe less than optimal.

We described a multivariate, multilevel, random effects, andcon rmatory factor model, based on con rmatory factor analysis orstructural equations modeling that is applicable to the analysis of datafrom idiographic assessment. The over-time relations between itemscores that are summarized in an intra-individual variance – covar-iance matrix that is analyzed with a con rmatory factor analysis. Foreach person, items with higher loadings indicate higher congruencefor that person and thus are relatively better measures of the targetconstruct — they have high congruence for that person. Importantly,

the magnitude of the loading for any particular item varies fromperson to person and, thus, is a random effect. This notion of therandom effect nature of the loading, the weight or importance of aparticular item, is formalized in the level 2 (across individuals) model.Thus, congruence is enhanced because the importance of differentitems varies across individuals.

Finally, we outline 15 steps in the development and evaluation of an idiographic assessment template, the selection of items from thetemplate for the individual, and validation of the template andinstrument.

References

American Psychological Association (2000). Diagnostic and statistical manual of mental

disorders (4th ed., text revision). Washington D.C.: Author.

190 S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179 –191

Page 13: Articulo CO-CO

8/4/2019 Articulo CO-CO

http://slidepdf.com/reader/full/articulo-co-co 13/13

Barton, S., Morley, S., Bloxham, G., Kitson, C., & Platts, S. (2005). Sentence completiontest for depression (SCD): An idiographic measure of depressive thinking. British Journal of Clinical Psychology, 44, 29−46.

Beck, J. S. (1995). Cognitive therapy: Basics & beyond. New York, NY: Guilford Press.Beck, A. T., & Emery, G. (1985). Anxiety disorders and phobias. New York: Basic Books.Burns, G. L., & Haynes, S. N. (2006). Clinical psychology: Construct validation with

multiple sources of information and multiple settings. In M. Eid & E. Diener (Eds.),Handbook of multimethod measurement in psychology. Washington, DC: AmericanPsychological Association.

Claes, L., Van Mechelen, I., & Vertommen, H. (2004). Assessment of situation-behaviorpro les and their guiding cognitive and affective processes: A case study from the

domain of aggressive behaviors. European Journal of Psychological Assessment , 20,216 −226.Clark,D. A.,Beck,A. T., & Alford, B. A. (1999).Scienti c foundations of cognitive theoryand

therapy of depression. Hoboken, NJ: John Wiley & Sons.Combs, D. R.,& Mueser,K. T.(2007). InM. Hersen, S.M. Turner& D. C. Beidel (Eds.), Adult

psychopathologyand diagnosis (pp. 234 −285). FifthEdition. NewYork:John Wiley &Sons, Inc.

Craske, M. G., & Waters, A. M. (2005). Panic disorder, phobias, and generalized anxietydisorder. Annual review of clinical psychology (pp. 167 −198). Palo Alto, CA: AnnualReviews.

Furr, M. R., & Bacharach, V. R. (2008). Psychometrics: An introduction. Thousand Oaks,CA: Sage.

Hammen, C. (2005).Stress anddepression. Annual review of clinicalpsychology (pp.293 −320).Palo Alto, CA: Annual Reviews.

Hayes, S. C., Barlow, D. H., & Nelson-Gray, R. O. (1999). The scientist practitioner: Researchand accountability in the age of managed care (2nd ed.). Boston: Allyn and Bacon.

Haynes, S. N. (1992). Models of causality in psychopathology: Toward synthetic, dynamic,and nonlinear models of causality in psychopathology. Boston: Allyn & Bacon.

Haynes, S. N., & Lench, H. C. (2003). Incremental validity of new clinical assessmentmeasures. Psychological Assessment , 15, 456−466.

Haynes, S. N., & O'Brien, W. H. (2000). Principles and practice of behavioral assessment.New York: Kluwer Academic/Plenum.

Haynes, S. N., Richard, D. C. S., & Kubany, E. (1995). Content validity in psychologicalassessment.A functional approachto concepts andmethods. Psychological Assessment ,7 ,238−247.

Haynes, S. N., Yoshioka, D., Kloezeman, K., & Bello, I. (2009). Clinical applications of behavioral assessment: Identifying and explaining behavior problems in clinicalassessment. In J. Butcher (Ed.), Oxford handbook of clinical assessment. Oxford:Oxford University Press.

Hersen, M. Turner, S. M. & Beidel, D. C. (Eds.). (2007). Adult psychopathology anddiagnosis (Fifth Edition). New York: John Wiley & Sons, Inc.

Hershberger, S. L. (1998). Dynamic factor analysis. In G. A. Marcoulides (Ed.), Modernmethods for business research (pp. 217 −249). Mahwah, NJ: Erlbaum.

Heyman, R. E., & Slep, A. M. (2004). Analogue behavioral observation. In S. N. Haynes &E. H. Heiby (Eds.), Comprehensive handbook of psychological assessmentBehavioralassessment , Vol. 3 (pp. 162 −180). New York: John Wiley & Sons, Inc.

Hoff,K. E.,Ervin, R. A.,& Friman, P. C. (2005). Re ning functionalbehavioral assessment:analyzing the separate and combined effects of hypothesized controlling variablesduring ongoing classroom routines. School Psychology Review, 34, 45−57.

Hokanson, J. E., Tate, R. L., Niu, X., Stader, S., & Flynn, H. (1994). Illustration of concomitanttimeseriesanalyses ina caseof somatoformdisorder. Cognitive TherapyandResearch ,18,413 −437.

Hopko, D. R., Bell, J. L., Armento, M. E., Hunt, M. K., & Lejuez, C. W. (2005). Behaviortherapy for depressed cancer patients in primary care. Psychotherapy: Theory,Research, Practice, Training , 42, 236−243.

Howard, G. S., & Meyers, P. R. (1990). Predicting human behavior: Comparing idiographic,nomothetic, and agentic methodologies. Journalof Counseling Psychology, 37 , 227−233.

Johnson, P. L., & O'Leary, K. D. (1996). Behavioral components of marital satisfaction: Anindividualized assessment approach. Journal of Consulting and Clinical Psychology , 64,417 −423.

Jones-Alexander, J., Blanchard, E. B., & Hickling, E. J. (2005). Psychophysiologicalassessment of youthful motor vehicle accident survivors. Applied Psychophysiologyand Biofeedback , 30, 115−121.

Kern, J. M. (1991). An evaluation of a novel role-play methodology: The standardizedidiographic approach. Behavior Therapy , 22, 13−29.

Kiresuk, T. J. Smith, A. & Cardillo, J. E. (Eds.). (1994). Goal attainment scaling:

Applications, theory, and measurement Hillsdale, NJ: Erlbaum.Lindauer, R. J., Van Meijel, E. P., Jalink, M., Olff, M., Carlier, I. V., & Gersons, B. P. (2006).Heart rate responsivity to script-driven imagery in posttraumatic stress disorder:Speci city of response and effects of psychotherapy. Psychosomatic Medicine , 68,33−40.

Little, R. J. A., & Rubin, D. B. (1987).Statistical analysis with missing data. New York:Wiley.

Mace,F. C.,Vollmer, T.R., Progar, P. R.,& Mace,A. B. (1998). Assessment andtreatmentof self-injury. InT. S Watson & F.M. Gresham (Eds.), Handbook of childbehaviortherapy(pp. 413 −430). Plenum Press: New York.

McFall, R. M. (2005). Theory and utility — key themes in evidence-based assessment:Comment on the special section. Psychological Assessment , 17 , 312−323.

Minke, K., & Haynes, S. N. (2003). Sampling issues. In J. C. Thomas & M. Hersen (Eds.),Understanding research in clinical and counseling psychology : Lawrence ErlbaumAssociates.

Molenaar, P. C. M. (1985). A dynamic factor model for the analysis of multivariate timeseries. Psychometrica , 50, 181−202.

Molenaar, P. C. M. (2004). A Manifesto on psychology as idiographic science: Bringing theperson back into scienti c psychology, this time forever. Measurement , 2, 201−218.

Molenaar, P. C. M. (2005). Rejoinder to Rogosa's commentary on “A Manifesto onPsychology as idiographic science ”. Measurement , 3, 116−119.

Molenaar, P. C. M., & Valsiner, J. (2005). How generalization works through the singlecase: A simple idiographic process analysis of an individual psychotherapy. Inter-

national Journal of Idiographic Science , 1 Retrieved April 20,2005 from http://www.valsiner.com/articles/molenvals.htmMorgan, S. L., & Winship, C. (2007). Counterfactuals and causal inference. Cambridge:

Cambridge University Press.Mumma, G. H. (2001). Increasing accuracy in clinical decision-making: Toward an

integration of nomothetic-aggregate and idiographic – idiographic approaches. TheBehavior Therapist , 24, 77−94.

Mumma, G. H. (2004). Validation of idiographic cognitive schema in cognitive caseformulations: An intra-individual idiographic approach. Psychological Assessment ,16 , 211−230.

Mumma, G.H. & Fluck, J. (2009). A clinician-friendly approach to the intraindividualempirical testingofa cognitive-behavioralcase formulation. Manuscriptunder review.

Mumma, G. H., & Mooney, S. R. (2007). Comparing the validity of alternative cognitivecase formulations: A latent variable, multivariate time series approach. CognitiveTherapy and Research , 31, 451−481.

Mumma, G. H., & Mooney, S. R. (2007). Incremental validity of cognitions in a clinicalcase formulation: An intraindividual test in a case example. Journal of Psycho- pathology and Behavioral Assessment , 29, 17−28.

Needleman, L. D. (1999). Cognitive case conceptualization: A guidebook for practitioners.Mahwah, NJ: Lawrence Erlbaum Associates Publishers.

Nesselroade, J. R., McArdle, J. J., Aggen, S. H., & Meyers, J. M. (2002). Dynamic factoranalysis models for representing process in multivariate time-series. In D. S.Moskowitz & S. L. Hershberger (Eds.), Modeling intraindividual variability withrepeated measures data (pp. 233 −265). Mahwah, NJ: Erlbaum.

Nunnally, J. C.,& Bernstein, I. H. (1994). Psychometric theory (3rdEd.).NewYork: McGraw-Hill.Orr, S. P., & Kaloupek, D. G. (1997). Psychophysiological assessment of posttraumatic

stress disorder. In J. P. Wilson & T. M. Keane (Eds.), Assessing psychological traumaand PTSD (pp. 69 −97). New York: The Guilford Press.

Persons, J. B., & Tompkins, M. A. (2007). Cognitive-behavioral case formulation. In T. T.Eells (Ed.), Handbook of psychotherapy case formulation. (2nd Ed) New York, NY:Guilford Press.

Piasecki, T. M., Hufford, M. R., Solhan, M., & Trull, T. J. (2007). Assessing clients in theirnatural environments with electronic diaries: Rationale, bene ts, limitations, andbarriers. Psychological Assessment , 19, 25−43.

Rosenzweig, S. (1986). Idiodynamics vis-a-vis psychology. American Psychologist , 41,241−245.

Shadish,W. R.,Cook,T. D.,& Campbell,D. T.(2001). Experimental and quasi-experimentaldesigns for generalized causal inference. Boston, MA: Houghton Mif in Company.

Sheeber, L. B., Davis, B., Leve, C., Hops, H., & Tildesley, E. (2007). Adolescents'relationshipswith their mothers and fathers:Associationswith depressive disorderand subdiagnostic symptomatology. Journal of Abnormal Psychology , 116 , 144−154.

Shiffman, S. (2007). Designing protocols for ecological momentary assessment. In A.Stone, S. Shiffman, A. A. Atienza & L. Nebeling (Eds.), The science of real-time datacapture: Self-reports in health research. New York: Oxford.

Shoda, Y.,Mischel,W.,& Wright, J. (1994). Intra-individual stability in the organization andpatterning of behavior: Incorporating psychological situations into the idiographicanalysis of personality. Journal of Personality and Social Psychology , 64, 674−687.

Smith, G. T., Fischer, S., & Fister, S. M. (2003). Incremental validity principles in testconstruction. Psychological Assessment , 15, 467−477.

Smith, P. N., & Mumma, G. H. (2006, August). Testing a cognitive case formulation in aclinical training setting. In G. H. Mumma (Ed.), Evaluating cognitive behavioral case formulation hypotheses using repeated measurement. Symposium presented at the114th Annual Convention of the American Psychological Association, New Orleans, LA .

Snyder, D. K., Heyman, R. E., & Haynes, S. N. (2008). Assessing couples. In N. James &Butcher (Eds.), Handbook of personality assessment. New York: Oxford UniversityPress Do not.

Solomon, A., Arnow, B. A., Gotlib,I. H., & Wind, B. (2003). Individualized measurementof irrational beliefs in remitted depressives. Journal of Clinical Psychology, 59,439 −455.

Thatcher, I., & Haynes, S. N. (2001). A multivariate time-series regression study of pain,depression symptoms, and social interaction in rheumatoid arthritis. International Journal of Clinical and Health Psychology, 1, 159−180.

Wenzel, A., Graff-Dolezal, J., Macho, M., & Brendle, J. R. (2005). Communication andsocial skills in socially anxious and nonanxious individuals in the context of romantic relationships. Behaviour Research and Therapy , 43, 505−519.

Wood, P., & Brown, D. (1994). The study of intraindividual differences by means of dynamic factor models: Rationale, implementation, and interpretation. Psycholo- gical Bulletin, 116 , 166 −186.

191S.N. Haynes et al. / Clinical Psychology Review 29 (2009) 179–191