7. Repeated Measures/Mixed ANOVA

Question 1

Benefits of repeated measures ANOVA

Answer 1

Sensitivity:​ unsystematic variance is reduced which means it is more sensitive to experimental effects.

Economy:​ less participants needed but need to be careful for fatigue effects

Question 2

One way repeated ANOVA

Answer 2

SST (total variance) - > SSB (between participant variation, individual differences?) + SSW (within-participant variation) ->SSM (effect of experiment) + SSR (error, variability not explained by experiment)

Problems: ​Same participants; scores across conditions correlate which violates assumption of independence. This means we must observe the assumption of sphericity.

Question 3

Assumption of sphericity

Answer 3

It states that the correlation between conditions should be the same. Specifically, the variance in differences between conditions should be the same. If not, adjust degrees of freedom. It is measured using maulchys test (if p is non-sig then satisfied).

Can use one of three corrections. Use Greenhouse-geisser unless estimate is more than .75, then use huynh-feldt. Can also either take the average of the two groups or use a Manova.

Can use post hoc tests to compare for differences between conditions, these are multiple t-tests with stricter alpha levels to control for familywise errors (e.g. bonferroni) or contrasts

Question 4

Two-way repeated ANOVA

Answer 4

This would have two iv’s (e.g. type of alcohol and imagery shown on the evaluation of drinks).

You look at main effects but also interaction.

Question 5

Three-way mixed ANOVA

Answer 5

Mixed ANOVA is when both repeated and between measures are used (for example the dating example, within subjects are the looks and personalities of the dates, between is the gender of participants). So at least one iv is within and one is between. In this example, the dv is the rating of the dates.

For running a mixed measures ANOVA, data for one participant (each combination of iv’s) must be in a single column. If it’s in multiple columns then you need to restructure data (see slides). This would mean creating a variable for each combination of the iv’s, with the dv being the value of each variable.

You would look assumption of sphericity for this to.

You look at main effects, two interactions and three way interaction.