2
$\begingroup$

I want to carry out a power analysis on a one group repeated measures experiment using G*power.

I have a group of subjects who tested a set of products. Each subject test one time the products. Hence, I have one observation per cell.

To test the product effect, I used a two anova model with subjects as random effect with proc glm in sas.

I would like to calculate the sample size for a given study having the same design. Given an estimation of effect size, alpha, Beta and an estimation of "a correlation among the repeated measures" could I use the ratio variance of subjets/ (variance of subjects + variance of error) derived from the two way anova model as an estimation of this correlation?

$\endgroup$

1 Answer 1

1
$\begingroup$

When you include subject as a random effect in ANOVA you assume that the subject effect and the product effect are additive. Have you thought about negative covariances? Maybe the more one likes red socks the less they like green ones...

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.