1 / 25

Chapter 7: The Experimental Research Strategy

2. Experiment: Characteristics. Manipulation of IVHold other vars constantParticipants in all conditions are equivalentPersonal attributes (on average)Any variables relating to the DVUsually done by random ASSIGNMENT to conditions(random selection is an external validity issue)Why?. 3. Statistics.

Leo
Download Presentation

Chapter 7: The Experimental Research Strategy

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


    1. 1 Chapter 7: The Experimental Research Strategy Manipulating the IV Controlling Extraneous Variance Holding Extraneous Vars Constant Between Subjects Designs Within Subjects Designs Multiple-Group Designs Quantitative IVs Qualitative IVs Factorial Designs Summary

    2. 2

    3. 3 Statistics Descriptive v. inferential Parametric Partition vars into ratio of treatment/error Non-parametric No assumptions about the distributions

    4. 4 Manipulation of IV Conditions of the IV Experimental and control conditions Comparison Conditions Additional Control and Comparison Conditions Hypothesis testing Ruling out specific alternative explanations Characteristics of a good manipulation Construct validity Reliability Strength Salience

    5. 5 Manipulation of IV Conditions of the IV Experimental and control conditions Equivalence of ? Allows you to rule out nonspecific treatment effects Any differences between the conditions other than treatment Similar to placebo effects Comparison Conditions How does comparison group differ from control? It doesn’t Additional Control and Comparison Conditions Hypothesis testing (Bransford &Johnson, ’72) Why three conditions? No context, context before, context after Ruling out specific alternative explanations (Alloy, Abramson, & Viscusi, ’81) added control conditions Neutral mood, role-play to mood state-> demand

    6. 6 Manipulation of IV (con’t) Characteristics of a good manipulation Construct validity Use manipulation check (e.g. Mood from essay writing) Debrief interview; include in DV; pilot testing Is it sensitive enough? Are Ps attending to IV? Reliability Automate instructions; detailed scripts Strength Realistic level (for external validity, and mundane realism), Salience Make sure they notice it

    7. 7 Manipulation of IV (con’t) Using multiple stimuli IV Stimulus: person, object, event Examples from your project? Use only one stimulus for a condition E.g. training program to increase cooperation What would possible stimuli be? Avoid confounding: stimulus person (multiple char) Physical char; personal char

    8. 8 Manipulations (con’t) Controlling Extraneous Variance External (keep environment; time same) Internal to P (more difficult) Random assignment Ps > conditions Use homogenous sample Repeated measures (within subjects) Between subjects designs To ensure group equivalence 1. Simple random assignment of Ps 2. Matched random assignment

    9. 9 Between-Subjects Designs Simple random assignment (most used) How does this help to ensure group equivalence? Individual differences (error variance) is randomly distributed across all conditions How does Kidd &Greenwald’s (’88) do this? What individual difference variable that may affect the outcome is randomly distributed across conditions? Memorization skill (does not differentially affect group means) Is it ok to use “quasi-random” assignment? What the hell is that?!!!!

    10. 10 Between-Subjects Designs If random assignment doesn’t guarantee group equivalence, what can help? (why doesn’t it?) Matched random assignment can! What are some Variables to match on? Categorical v. continuous vars Which ones are more difficult to match on? Compare gender and IQ Which need a pretest? Any downside to pretesting? Does the pretest variable need to be related to the DV?

    11. 11 Within-Subjects Designs Ps participate in each condition Advantages Control individual differences (Perfect match) What does this do? Reduce error (random) variance Fewer Ps needed (increased power) Disadvantages Order effects Practice effects Carryover Sensitization E.g. Wexley et al. (’72) what was the problem? Demand effects

    12. 12 Within-Ss Controls Order effects Counterbalancing Latin Square Basic v. balanced What’s the difference? = Sequence v. order What’s a washout period? Differential order effect (Table 7-4) Sensitization / demand characteristics Don’t use repeated measures Order effects can be of theoretical interest Build into the experiment

    13. 13 Multiple Group Designs Quantitative IVs Linear relationships What is an e.g. of a linear IV for your project? Positive / negative / curvilinear? What is the minimum levels necessary for quantitative? Why? 3… 2 points can only define a straight line DeJong et al. (’76); Feldman & Rosen (’78); Whitley (’82) What happened? Qualitative IVs Give an e.g. of a qualitative IV for your project

    14. 14 Multiple-Group Designs Interpreting the Results One way ANOVA Post hoc or Contrasts (Planned comparisons) What’s the difference? A priori (Before=contrasts) v. Post hoc (After) Compare omnibus F with focused F tests What is the benefit of a priori?

    15. 15 INCREASING THE NUMBER OF LEVELS OF AN INDEPENDENT VARIABLE Provides more information about the relationship than a two level design Curvilinear Relationship Inverted-U Comparing Two or More Groups I.E. How dogs, cats, and birds as opposed to dogs alone have beneficial effects on nursing home residents

    16. 16 Factorial Designs Nature of Factorial Designs Describing them 2X2 (how many factors? Levels? Conditions? 2 factors, 2 levels each = 4 conditions 4X2 2 factors, 4 and 2 levels= 8 conditions 2X3X2 3 factors, 2, 3, & 2 levels =12 conditions Information provided Main effects (how many in each example above?) Interactions (how many 2 way; three way?) What did Platz & Hosch (’88) find? What caused the interaction to occur?

    17. 17 Factorial Designs Displaying interactions Which is clearer? Line or bar graph? (fig 7-5) Convert from table of means to graph (fig 7-6, p. 208 -209) Interpreting interactions Main effects, interactions, both? Theory driven? (a priori v. post hoc)

    18. 18 Factorial Designs: Forms Forms of Factorial Designs Between & Within-Subjects Designs Between: Each subject participates in only one condition Within: Each subject participates in all conditions Mixed: Each subject participates in more than one condition Platz & Hosch (’88) Store clerk (between) could it be within? Customer (within) could it be between? Manipulated & Measured IVs Manipulated IV: true experimental design Measured IV: correlational aspect of design Caveat: Don’t dichotomize when not needed

    19. 19 Factorial Designs: Forms Design Complexity Factors and levels (already discussed) How many Ps needed for Between design With 10 per condition? 2X3? 60 Ps 3X4X2? 240 Ps

    20. 20 INCREASING THE NUMBER OF INDEPENDENT VARIABLES: FACTORIAL DESIGNS N = 40 © 2007 The McGraw-Hill Companies, Inc.N = 40 © 2007 The McGraw-Hill Companies, Inc.

    21. 21 INCREASING THE NUMBER OF INDEPENDENT VARIABLES: FACTORIAL DESIGNS n=10 © 2007 The McGraw-Hill Companies, Inc.n=10 © 2007 The McGraw-Hill Companies, Inc.

    22. 22 INCREASING THE NUMBER OF INDEPENDENT VARIABLES: FACTORIAL DESIGNS n=20 © 2007 The McGraw-Hill Companies, Inc. n=20 © 2007 The McGraw-Hill Companies, Inc.

    23. 23 Uses of Factorial Designs Testing Moderator Hypotheses Moderator: changes the effects of IV E.g. Platz & Hosch (’88) race of clerk Use of ANCOVA & MR Detecting Order Effects Table 7-6 Top: main for condition; no main for order; no interaction Middle: main for condition; no main for order; interaction Bottom: main for condition & order; interaction

    24. 24 Blocking on Extraneous Vars Including it as an IV Ps are grouped on extraneous var and tested by ANOVA as a factorial Blocking reduces the error term (fig 7-9) Caveat: Remember that the blocking var cannot be explained as cause

    25. 25 Experimental Strategy: Summary Manipulating the IV Controlling Extraneous Variance Holding Extraneous Vars Constant Between Subjects Designs Within Subjects Designs Multiple-Group Designs Quantitative IVs Qualitative IVs Factorial Designs Summary

More Related