I'm using G*Power to work out how many participants I need. I'll be using a MANOVA, I have two independent variables (Male and Female) and they will be answering 13 different questionnaires so 13 dependent variables. I put this data into G*Power and its telling me I only need 14 participants but I think that is too small?
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
2
If you choose as test family F tests and as statistical test MANOVA: Global effects. You can choose as Type of power analyis A priori: Compute required sample size - given alpha, power, and effects size. Depending on the Effect size, alpha and power you would like to achieve you obtain the required sample size (with 2 for the number of groups, and 13 for the response variables). Of course the effect size, etc. are up to you to decide and should be based, among other things, on theory!
-
$\begingroup$ Hi all. Just asked a similar question (stats.stackexchange.com/q/263308/64052) about a MANCOVA and would appreciate your help. $\endgroup$ Commented Feb 21, 2017 at 23:45
-
$\begingroup$ Remember that the 438 means at least 219 (half) in each of the two groups. $\endgroup$– bashinskCommented Oct 14, 2018 at 16:19
$\begingroup$
$\endgroup$
G*Power defines a medium-sized effect with the same value of f-squared (.0625) used for a medium effect in univariate ANOVA. However, Cohen suggested different benchmarks for MANOVA, with a medium-sized effect having an f-squared of .15. See Steyn, H. S., Jr. & Ellis, S. M. (2009). Estimating an effect size in one-way multivariate analysis of variance (MANOVA). Multivariate Behavioral Research, 44(1), 106-129.
