Questions about MANOVA and ANOVA??????

#1
I have a BUNCH of questions related to a study I posted few weeks ago. ANY help on ANY of them would really really be appreciated!!!!

1) Can I do MANOVA or ANOVA with 3 groups? If so, group would be an independent variable, correct? Would say I was doing a between group analysis or a across group analysis?

2) If in the literature previous exploratory and confirmatory factor analysis the 8 sub-tests have loaded on 4 factors (each consisting of 2 sub-tests). Each of these 4 factors would be a dependent variable, right?

3) I want to compare STM to each of the 4 factors between/across groups. Can I have 2 independent variables in MANOVA or ANOVA? My second independent variable would be STM, right?

4) If I do MANOVA:

a) I would be comparing the 2 independent variables (STM between/across groups) to each of the 4 factors simultaneously. This would let me know the correlation between STM and each dependent variable, accounting for shared variance (right term?) between dependent variables (STM and Factor 1, STM and Factor 2, etc.)

b) Would it tell me if the is a significant difference between the groups on the correlation between SEM and each of the 4 factors?

c) Would it tell me what amount of variance is shared between these dependent variables (how much variance is shared between Factor 1 and 2, Factor 1 and 3, Factor 2 and 3, etc?)?
OR
Is this actually SEM and too lofty a goal because of my small N. If so, is there a way to do this not as SEM?
OR
Is the just the covariance matrix?

d) Would if tell me if the is a significant difference between the groups on the amount of variance shared?

5) Since my n is small, if MANOVA does not work, can I do post hoc do ANOVA? Is this poor research? If this is not ok, and I think this will happen, should I just make my hypothesis represent that my analysis will be ANOVA, not MANOVA?

6) If I do ANOVA:

a) I would be comparing the 2 independent variables (STM between/across groups) to each of the 4 factors independently. This would let me know the correlation between STM and each dependent variable, NOT accounting for shared variance (right term?) between dependent variables (Analysis 1: STM and Factor 1, Analysis 2: STM and Factor 2, etc.)

b) Would it tell me if the is a significant difference between the groups on the correlation between SEM and each of the 4 factors independently?

c) I would have a higher Type 1 error rate and lower power?

d) Should I use something like a Bonferoni (sorry about spelling) correction? This would control for a family wise (or is it wide?) error rate, right?

7) With either analysis, could I look at the sub-tests and STM between/across the 3 groups?

8) If I have separated the groups with other measures, do I need to begin all of the analysis making sure the groups are significantly different on the performance on STM, Factor 1, Factor 2, etc. to decide what to include in the MANOVA or ANOVA?

Thanks so much to anyone that can help. I know I am basically asking for a lesson on multivariate analysis, but I have had trouble finding someone to talk to about the details and have trouble finding exactly what I need in books. If anyone can direct me to any studies or examples of studies that have a similar analysis, or stats books, that would also be so amazing.
 

pmcc

New Member
#2
I am reading Applied Multivariate Techniques by Subhash Sharma (John Wiley). I find it very well structured, but I don't know whether it would help you on MANOVA. Perhaps you might be able to find a preview of the book online.