Upload
dimitri-carron
View
218
Download
4
Embed Size (px)
Citation preview
Types of Experimental Designs• Non-experimental
– Post-test only– Pre-post test
• Experimental– Pre & post-test with control group– Pre & multiple post-test with control group – After-only control group– Solomon four-group
• Quasi– Non-equivalent control group– Simple time-series– Multiple time series
Laboratory vs. Field Experiments
• Lab– Artificial (Low external validity)– Control of variables (High internal validity)– Good for theory building and process-level
understanding
• Field– Realistic (High external validity)– Less control (Low internal validity)– Good for program evaluation
Non-Experimental
• No control group
• No measurement of dependent variable before intervention
• Referred to as quasi-experiment in Sekaran
Non-Experimental• Post-test only
– Referred to as quasi-experiment in Sekaran– Measurement taken only after intervention– No baseline measure taken– Hard to determine if change
• has occurred • is due to intervention
– Good for studying nature of intervention– Useful for evaluating participant learning after a
training session
Intervention Post-measure
Non-Experimental• Pre-test/post-test
– Referred to as quasi-experiment in Sekaran
– Change inferred from differences in pre vs. post intervention measures
Intervention Post-measurePre-measure
Non-Experimental• Pre-test/post-test
– Is affected by testing, instrumentation, history, time, maturation, & statistical regression effects
– Use internal referencing data to test whether only relevant var changed and irrelevant var did not change following intervention
Intervention Post-measurePre-measure
Experimental
• Control group
• Random assignment of participants to conditions– Is the study described in the ethnic identity
paper (Kim & Gelfand) an experiment? Why or Why not?
Experimental• Pre-test/post-test w/control group
Intervention Post-measurePre-measure
Post-measurePre-measure
Experimental group
Control group
• Subject to mortality threats
Experimental• Pre-test/post-test w/control group (omnibus
test= repeated measures anova)– Experimental group should be ‘better’ than
control group on post-test measures (t-test independent groups)
– Experimental group scores should “improve” from pre to post-test (t-test paired)
– Control group scores should not change from pre to post-test (t-test paired)
– Should be no group differences on pre-test measures esp on matching variables if random assignment was not possible (t-test independent groups)
• Pre-test & Post-test Control Group with multiple post-test measures
Experimental
Intervention Post-measure 1,2,3Pre-measure
Post-measure 1,2,3Pre-measure
Experimental group
Control group
– Tests long and short term effects of intervention (e.g., training)
• Pre-test & Post-test Control Group with multiple post-test measures (adapted to org’n stakeholders’ needs)
Experimental
Intervention Post-measure 1Pre-measure
Experimental group
Control group
Control Post-measure 2
Control Post-measure 1Pre-measure
InterventionPost-measure 2
Experimental• After only control group
Intervention Post-measure
Post-measure
Experimental group
Control group
• Eliminates testing effects found in pre-test post-test control group design
• Classified as quasi-experimental in Sekaran • Referred to as post-tests only with experimental &
control groups in Sekaran• Subject to Selection & mortality threats
• Solomon Four-group design
Experimental
Intervention Post-measurePre-measure
Post-measurePre-measure
Experimental group
Control group
Intervention Post-measure
Post-measure
Experimental group
Control group
TABLE 7.2Copyright © 2003 John Wiley & Sons, Inc. Sekaran/RESEARCH 4E
Types of Experimental Designs
• Experimental– Pre & post-test with control group– Pre & multiple post-test with control group – After-only control group– Solomon four-group
• Quasi– Non-equivalent control group– Simple time-series– Multiple time series
Quasi-Experimental
• Non-equivalent control group– E.g., training conditions assigned to pre-
existing groups– Gather pre-intervention data on important
control variables.– Use pre-measure as a statistical control in
the analyses for differences in post-intervention measures
• Simple time-series design
Quasi-Experimental
Intervention Post-measure 1,2,3Pre-measure 1, 2, 3
• Pre-measures should not be different from each other
• Post measures should not be different from each other
• Pre vs. post measures should be different from each other
• Multiple time-series design
Quasi-Experimental
Intervention Post-measure 1,2,3Pre-measure 1, 2, 3
Post-measure 1,2,3Pre-measure 1, 2, 3Control group
Experimental group
• Same conditions as simple time series design plus..
• Pre and post measures of control group should not differ from each other
• Limitations: Testing effects, instrumentation problems
Types of Experimental Designs• Non-experimental
– Post-test only– Pre-post test
• Experimental– Pre & post-test with control group– Pre & multiple post-test with control group – After-only control group– Solomon four-group
• Quasi– Non-equivalent control group– Simple time-series– Multiple time series