Repeated Measures & Three Factor ANOVA

Question 1

Subject Mean

Answer 1

Reduces error term in repeated measures design
What a participant’s score is if you average across the independent variable
Baseline individual differences
Factored out of the error term (removing known variability)
Increases power of design

Question 2

Within Subject Variability - Repeated Measures

Answer 2

Within subject variability is not all error anymore
Remove treatment effect (e.g. time) from error term (Degree to which the effect varies as a function of subject)
Shrinks the error term, increases power

Question 3

Reporting Results For Repeated Measures ANOVA

Answer 3

We don’t report between subject, also don’t analyze it

Question 4

ANOVA Assumptions

Answer 4

Observations normally distributed within each population
Population variances are equal (e.g. homogeneity of variance or homoscedasticity)
Observations are independent (not for RM)
Compound symmetry/sphericity of the covariance matrix (homogeneity of covariance)

Question 5

Homogeneity of Covariance

Answer 5

Compound symmetry/sphericity of the covariance matrix
Assumes the relations between each level of IV are relatively homogeneous
Assume correlation between time 1 and time 2 is comparable to correlation between time 2 and time 3, etc.
Need to worry about this because violations affect chance of type 1 error
SPSS uses Mauchly test of Sphericity

Question 6

Significant Mauchly Test

Answer 6

Corrects using a Greenhouse/Geisser Correction (adjusts degrees of freedom)
Makes df smaller, which makes test more conservative

Question 7

RM Multiple Comparisons With Few Means

Answer 7

T test with Bonferroni corrections would be simplest
Limit to important comparisons

Question 8

RM Multiple Comparisons With Many Means

Answer 8

Can use N-K or Tukey, but must do by hand
Paired samples t-test

Question 9

Cohens D

Answer 9

Cohen’s d for effect size just between two means
MS error = pooled variance, so sqrt of MSerror is pooled standard deviation