Minimum Sample Size for MANOVA: Help!

P

Quick questions for any stats gods:

I'm trying to determine if my sample size for a MANOVA is acceptable. I've read in several textbooks that the you need to 'have more cases in each cell than you have DVs'. However, I'm a bit stuck about what this means!

I have done multiple MANOVAs, and some have as many as 6 DVs. Now, does that mean I have to have more participants each each cell than DVs (i.e. more than 6 in each group?) OR does it mean that as I have 6 DVs and 3 IVs (6 x 3) I need 18 cases (i.e participants in each group?)

I'm a bit concerned as I have 20 in one group, 30 in another but just 11 or 12 in another group.

Any help would be much appreciated as I'm getting a bit stressed out :-)

Avatar for sneaks

======= Date Modified 11 Oct 2011 11:34:17 =======
I'd take it as more than 6 in each cell. I'd then stick in a load of references to support you doing this. I'd also bootstrap the analysis you can do this on SPSS 19 and if this is for your PhD, I'd definately do this with difference sizes in each group.

E.g. MANOVA was used with X Y and Z as DVs and A as an IV, this methods was chosen because it provides enhanced power (Refs), the sample size was also adequate for the analysis (ref ref)

P

Thanks for such a quick reply Sneaks. So you mean more than 6 in each cell (i.e group?), so according to those references I should be okay?

Also, sorry to be dim here, but what do you mean by bootstrap?

Thanks for your help!

Avatar for sneaks

Yeah, I'd go with 6 in each cell, but then, for I'd go with whatever works and then reference it anyway :p

Bootstrap - on later versions of SPSS there is a button on the MANOVA pop up - near the usual ones, that says 'bootstrap' - have a quick google, its basically a resampling technique, but enhances the robustness of the test.

P

Thanks ever so much, Sneaks!

20751