Psychology 513: Advanced Research Methods

1
Psychology 513 Advanced Research Methods

• Repeated Measures Designs

2
Overview

• Same subjects in each condition
• Between subjects ANOVA error term contained
  individual differences error
• Within-subjects design can separate out the
  individual differences so error term will be much
  smaller

3
Figure 14.1 (p. 473)
4
The Structural Model

• Model 1 (Additive Model)
• Xij m pi tj eij
• Model 2 (Nonadditive Model)
• Xij m pi tj ptij eij
• Includes subject x treatment interaction

5
Expected Mean Squares

• Model 1
• Subjects s2e ks2p
• Treatments s2e ns2t
• Error s2e
• Model 2
• Subjects s2e ks2p
• Treatments s2e s2pt ns2t
• Error s2e s2pt

6
Assumption of Sphericity

• Homogeneity of variances of differences scores -
  violations lead to inflated Type 1 error rates
• Sufficient, but not necessary, conditions
• Additivity
• Compound symmetry

7
What do you do if you have nonsphericity?

• Multivariate ANOVA (MANOVA)
• Adjusted degrees of freedom
• Greenhouse Geisser more conservative
• Huynh Feldt- more liberal
• Planned contrasts
• Sphericity doesnt apply when have 1 df

8
Multiple Repeated Measures

• Choice of an appropriate error term
• Each effect has a different error term
• Use General Linear Models to perform analysis

9
2 x 3 within-subjects design

• Main effect of A error term is S/A
• Main effect of B error term is S/B
• Interaction between A B error term is S/AB