Single Variable, Independentgroups Designs

1
Single Variable,Independent-groups Designs

• Chapter 10
• Psyc 50

2
Experimental Designs

• Test one or more hypotheses about causal effects
  of the independent variable(s) (IV)
• Include at least two levels of the IV
• Randomly assign participants to conditions
• Include specific procedures for testing
  hypotheses
• Include control for the major threats to internal
  validity

3
Single variable, independent-groups designs

• Randomized, posttest-only, control-group design
• Randomized, pretest-posttest, control-group
  design
• Multilevel, completely randomized,
  between-subjects designs
• Solomons four-group designs

4
Randomized, Posttest-Only,Control-Group Design

• DesignR Group A Treatment
  PosttestR Group B No Treatment Posttest
• Key element is random assignment to groups!
• Random assignment controls for selection
• Other confounding variables are controlled by
  comparing the treatment and no treatment groups

5
Randomized, Pretest-Posttest,Control-Group Design

• DesignR Group A Pretest Treatment
  PosttestR Group B Pretest No
  Treatment Posttest
• Adding a pretest allows us to quantify the amount
  of change following treatment
• Also allows us to verify that the groups were
  equal initially
• A strong basic research design, with excellent
  control over confounding

6
Multilevel, Randomized,Between-Subjects Design

• DesignR Group 1 Pretest Treatment 1
  PosttestR Group 2 Pretest
  Treatment 2 Posttest
• R Group N Pretest Treatment N
  Posttest
• May or may not include a pretest
• Multi-group extension of the basic experimental
  design

7
Solomons Four-Group Design

• DesignR Group A Pretest Treatment
  PosttestR Group B Pretest No
  Treatment PosttestR Group C
  Treatment PosttestR Group D
  No Treatment Posttest
• Combines two basic experimental designs
• Allows us to assess whether there is an
  interaction between the treatment and the pretest

8
Statistical Analysis Issues

• If the data are nominal, use chi-square
• If the data are ordinal, use the Mann-Whitney
  U-test
• If the data are interval or ratio
• If there are only two groups, a t-test of the
  posttest measures will test the hypothesis
• More complex designs will require an ANOVA

9
Data analysis for a two-group experiment

• One population, two samples
• Random error
• Affects individual scores
• May cause group means to differ

10
Data analysis for a two-group experiment

• Within group variability due to random error
• Between group variability due to random error
  treatment effect

11
Data analysis for a two-group experiment

• t-test is the difference between two groups
  larger than would be expected by random error
  alone?
• Group 1 mean Group 2 mean
• Standard error of the difference

t
12
Data analysis for a multi-group experiment

• Within-groups (error) variance
• Between-groups (treatment) variance

13
Data analysis for a multi-group experiment

• F MSB MS Treatment MST
• MSW MS Error MSE
• F Random Error Possible Treatment Effect
• Random Error

14
Example

• Effect of alcohol on hand-eye coordination
• Three levels of IV
• 0 shots 3 shots 6 shots
• DV number of errors on simulated driving task

15
Possible outcomes

• A.
• 0 shots 3 mistakes
• 3 shots 10 mistakes
• 6 shots 10 mistakes
• B.
• 0 shots 3 mistakes
• 3 shots 3 mistakes
• 6 shots 10 mistakes
• C.
• 0 shots 3 mistakes
• 3 shots 6 mistakes
• 6 shots 9 mistakes

Each of these could provide a significant
over-all F-test
16
Specific Means Comparisons

• A significant F-test means that at least one
  group is significantly different from at least
  one other group
• If you have more than two groups, you have to do
  follow-up tests (contrasts) to see which groups
  differ