High Speed Hupotheses

Embed Size (px)

Citation preview

  • 8/3/2019 High Speed Hupotheses

    1/3

    High-speed hypotheses!

    Lawrence M. Lesser

    The University of Texas at El Paso, Texas, USA.

    e-mail: [email protected]

    Abstract This article presents engaging interactive hypothesis tests which can be conducted

    with students very efficiently.

    Keywords: Teaching; Hypothesis tests; P-value.

    Introduction

    With class time at a premium, it helps to have

    handy a repertoire of engaging classroom-tested

    activities or experiments that can each be done

    within 5 minutes from collection of data to

    computation of the p-value! Such activities

    support the American Statistical Associations

    (ASA; 2010) recommendations to use real

    data and foster active learning in the class-

    room, illustrate the statistical problem solving

    process (ASA 2007) and likely tap the potential

    benefits of using fun in the statistics classroom

    (Lesser and Pearl 2008). A representative

    variety of common basic hypothesis tests are

    spanned with the activities which follow.

    Activity one: who has ESP?

    Background: Because it is discussed in popular

    culture, students are already familiar with and

    interested in the idea of having extra-sensory

    perception (ESP). Some textbooks (e.g. Utts

    2005) have used it in case study examples. ESP

    experiments commonly involve predictions from

    a smaller set of items (with maybe 4 or 5

    choices), but this can be modified for pedagogi-cal purposes, as it is here.

    Topic: (One-tailed) hypothesis test for a pro-

    portion p

    Procedure: You bring in a standard 52-card

    deck of cards (make sure students are familiar

    with the breakdown of suits and values on the

    cards) and have students commit in writing

    their predictions for, say, a dozen cards about

    to be drawn with replacement. Under the null

    hypothesis of no ESP, the probability of a

    correct prediction (i.e. of both suit and denomi-

    nation) would be 1/52, or about 1.9%. While

    you draw the cards, every student in the room

    is participating by independently trying to make

    a correct prediction. Because almost all stu-dents will experience no more than two suc-

    cesses, it is very quick to tabulate the results

    of the hypothesis test for these cases. When

    asked in advance Do any of you have ESP?,

    no student seriously answers Yes, so it can

    be noted that by chance alone, a sufficiently

    large class can be expected to have at least

    one student who obtains a personal p-value of

    less than 5%, therefore falsely rejecting the

    null hypothesis.

    Teachers interested in adding a magic trick

    component (e.g. using marked cards) can ensure

    obtaining a significant p-value by being (almost

    surely) the only one in the room to make all

    predictions correctly or (to make it not quite so

    suspicious) even to make most of the predictions

    correctly (by deliberately missing a few). Lesser

    and Glickman (2009) offer related discussion on

    the use of magic in statistics. Teachers wanting to

    vary the success probabilities involved could have

    students predict only the suit or only the denomi-

    nation. A website students can use to further

    explore and assess their psychic abilities

    (to predict cards, pictures, locations, lottery

    numbers, etc.) is http://www.gotpsi.org.

    Activity two: take your seats

    Background: There is some literature about

    where students prefer to sit in various types of

    classrooms (e.g. Kaya and Burgess 2007) and

    gender is a variable that is considered of poten-

    tial importance in this context. For this activity,

    we assume that the classroom seating arrange-

    ment is a rectangle made up of rows of individual

    seats, but modifications can be made for other

    arrangements.

    Practical activities

    2011 The AuthorTeaching Statistics 2011 Teaching Statistics Trust, 34, 1, pp 101210

  • 8/3/2019 High Speed Hupotheses

    2/3

    Topic: Test of the equality of two proportions (or

    done as a chi-squared test of independence)

    Procedure: Walk into the classroom where stu-

    dents have already freely chosen their seats.

    Have the class agree on an operational definition

    of front half of the room and back half of theroom, for the part of the room where students

    are sitting. The class can fill in the table below in

    less than 1 minute and then go on to test the null

    hypothesis that males and females are equally

    likely to sit in the front half of the room. The

    teacher could have students identify the two

    populations in this activity and then discuss

    whether or not the males and females in the

    room can be considered as random samples

    from those respective populations.

    Activity three: hold a note

    Background: Some published literature

    (e.g. Cheng 1991) suggests that you can sing

    more fully in a standing position than in a sit-

    ting position, while other literature suggeststhat the more important consideration is sim-

    ply good posture whether you are sitting or

    standing.

    Topic: Testing the equality of two means

    Procedure: Each student flips a coin to

    determine whether he/she will be in the sitting

    (tails) or standing (heads) group. An online

    stopwatch (e.g. http://www.shodor.org/

    interactivate/activities/Stopwatch/) is projected

    onto a screen where it is within view of every-

    one. After a 3-2-1 countdown by the teacher, the

    stopwatch will be started and everyone in the

    room will start to sing a note. (For consistency,

    suggest that everyone use a common syllable

    such as ohhhhhhh; also suggest to everyone to

    choose a pitch in the middle of their singing

    range, which should be easier to maintain for the

    duration of their singing). Each student is asked

    to watch the stopwatch and notice the amount of

    elapsed time at the instant he/she ran out of air

    and was unable to continue singing. Almost

    everyone will be done after about 18 seconds, so

    the total time needed is very minimal. The

    teacher now displays a spreadsheet or software

    package for data analysis. The data can be

    entered either by having students one at a time

    call out their time (and whether they were sitting

    or standing) or, in case more privacy is desired,

    by having students fill out and turn in a simple

    slip of paper without their name on it that indi-

    cates the time duration (in seconds) andwhether they were sitting or standing.

    To perform the test of means of the two

    independent samples, students will need to

    agree on whether to have the software do a t

    test or a z test, and whether to assume the

    unknown variances are equal. Beyond the cal-

    culations, students can also be asked to do the

    following:

    (1) discuss whether this experiment is best

    classified as single-blind, double-blind or nei-

    ther; (2) discuss what would be changed to

    have each person try both the sitting and the

    standing conditions; (3) consider a third condi-

    tion (e.g. singing with arms held straight up

    towards the ceiling, or maybe singing right

    after doing 20 jumping jacks) which would

    make this activity a vehicle to test the equality

    of more than two means (i.e. to use analysis of

    variance).

    Even though singing can be a personal form

    of expression, the beauty of this activity is that

    it simply is a single note (so that the ability to

    keep rhythmic time, for example, is not

    needed) and that everyone is doing it at the

    same time and so no one will be able to easilyhear anyone (or hear when they drop out)

    except for the very last couple of people, and

    by that naked end of the window, students

    have already moved beyond any initial moment

    of self-consciousness. To make the activity feel

    safe and fun, I demonstrate the procedure

    by holding a note all by myself first while I run

    the stopwatch, then I point out that each voice

    will be camouflaged by doing it all together,

    and I also mention that anyone who feels

    uncomfortable for any reason is not required to

    participate. To the extent that humming may

    feel safer than singing, this activity works

    just as well having each student humming

    (with closed lips, mmmmmm). In several

    implementations of this activity, I have not

    noticed any students reticence or non-

    participation. If anything, there seems to be a

    positive communal experience about doing

    something this unusual in statistics class in

    general, as well as the particular pleasantness

    of being enveloped by the sound vibrations of

    everyones simultaneous notes and how they

    collectively and spontaneously formed interest-

    ing harmonies.

    Number

    of males

    Number

    of females

    Front half of the classroom

    Back half of the classroom

    High-speed hypotheses! 11

    2011 The AuthorTeaching Statistics 2011 Teaching Statistics Trust, 34, 1, pp 1012

  • 8/3/2019 High Speed Hupotheses

    3/3

    Activity four: which hand is quicker?

    Background: Anderson-Cook and Sundar (2001)

    describe a class demonstration to compare the

    reaction times of ones dominant and non-

    dominant hands. In the demonstration, there isa 3-2-1 countdown and then one of nine boxes

    lights up and the time (to the hundredths of a

    second) is recorded for how long it takes to

    move the computer mouse (with a designated

    hand) over to that highlighted box and click on

    it. As time permits before the demonstration,

    teachers can first discuss the principles of

    experimental design and critique the flawed

    first draft of the experimental design offered in

    that article. If time is limited, however, teachers

    can just go to and use the final version of the

    applet at http://www.amstat.org/publications/

    jse/java/v9n1/anderson-cook/GoodExpDesign

    Applet.html.

    Topic: Test of means from two dependent

    samples

    Procedure: Teacher projects onto a screen the

    website at the URL just mentioned and person-

    ally takes the test with each hand to demon-

    strate to the class how the applet works,

    explaining what dominant and non-dominant

    mean and how the traffic light countdown

    graphic and Reset! and Start! buttons work.

    Then, a reasonable number of students (as few

    as 5 or 6 is sufficient to illustrate the concept,but more can be allowed if there is time and

    interest) each come up one at a time to try the

    applet. Each data observation takes just a

    second (literally), so the entire data collection

    can go very quickly. Then the Compile Info!

    button can be hit to reveal the t value of the

    paired t test.

    References

    American Statistical Association (2007). Guide-

    lines for Assessment and Instruction in Statis-

    tics Education (GAISE) Report: A Pre-K-12

    Curriculum Framework. Alexandria, VA:American Statistical Association. http://

    www.amstat.org/education/gaise/.

    American Statistical Association (2010). Guide-

    lines for Assessment and Instruction in Statis-

    tics Education: College Report. Alexandria, VA:

    American Statistical Association. http://

    www.amstat.org/education/gaise/.

    Anderson-Cook, C.M. and Sundar, D.-R. (2001).

    An active learning in-class demonstration of

    good experimental design. Journal of Statistics

    Education, 9(1), http://www.amstat.org/

    publications/jse/v9n1/anderson-cook.html.Cheng, S.C.-T. (1991). The Tao of Voice: A New

    East-West Approach to Transforming the

    Singing and Speaking Voice. Rochester, VT:

    Destiny Books.

    Kaya, N. and Burgess, B. (2007). Territoriality:

    Seat preferences in different types of class-

    room arrangements. Environment and Behav-

    ior, 39(6), 859876.

    Lesser, L.M. and Glickman, M. E. (2009). Using

    magic in the teaching of probability and sta-

    tistics. Model Assisted Statistics and Applica-

    tions, 4(4), 265274.

    Lesser, L.M. and Pearl, D.K. (2008). Functionalfun in statistics teaching: Resources, research,

    and recommendations. Journal of Statistics

    Education, 16(3), 111. http://www.amstat.

    org/publications/jse/v16n3/lesser.pdf.

    Utts, J.M. (2005). Seeing Through Statistics (3rd

    edn). Belmont, CA: Thomson Brooks/Cole.

    12 Lawrence M. Lesser

    2011 The AuthorTeaching Statistics 2011 Teaching Statistics Trust, 34, 1, pp 1012