Regression following MANOVA results

#1
Hi Stats Gurus,

This question is about a regression analysis that I want to run, but I am not too familiar with this type of analysis.

I have conducted MANOVAs to isolate variables that are significant out of the ones that were theoretically relevant for my analysis. I would now like to conduct regression analyses to see if these variables predict a different outcome, but I am wondering if I should include all of my theoretically-relevant variables that I initially used for the MANOVAs or whether it would make more sense to only include the variables that came out significant.

J.
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
So is this a completely different outcome in the second analyses or is it partially related to the first. When moving forward you still need to think about false discovery rate and trying to maintain your original level of significance (alpha), because after awhile you may find significance by chance (e.g., 1/20, type I error for an alpha=0.05).
 
#3
So is this a completely different outcome in the second analyses or is it partially related to the first. When moving forward you still need to think about false discovery rate and trying to maintain your original level of significance (alpha), because after awhile you may find significance by chance (e.g., 1/20, type I error for an alpha=0.05).
Yes, I agree regarding your comment about type 1 error.

The outcome in my regression analysis is partially related to the first outcome, for which some predictors came out statistically significant. In that case would that justify me only using those predictors that came out significant?
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
This all depends on the purpose of the analyses. Is it a fish expedition, are their actual hypotheses and a protocol for the analytics, how closely are the outcomes related. If they are only partial related then their predictors could vary.
 
#5
This all depends on the purpose of the analyses. Is it a fish expedition, are their actual hypotheses and a protocol for the analytics, how closely are the outcomes related. If they are only partial related than their predictors could vary.
Gotcha. Thanks for your insight!

J.