2
$\begingroup$

Suppose I have two groups and several dependant variables. I would like to show statistically whether the two groups are different or not, based on the combination of these dependant variables. The dependant variables will be correlated to some extent.

When doing t-test or ANOVA each of my dependant variables are almost significant based on my chosen p-value cut-off.

I would like to harness the power of all of my dependant variables together and get a single p-value out from the comparison.

I could do this by a repeated measures ANOVA but perhaps this wouldn't be appropriate since it would be more accurate to say that my measurements are of related but seperate dependant variables rather than an individual dependant variable with repeated measurements. It would probably be more appropriate to model the situation with a MANOVA, but with that test I don't think I can get a single value of significance to summarise over all dependant variables?

$\endgroup$
3
  • 3
    $\begingroup$ If you do a MANOVA on two groups you should indeed be able to get a single p-value. However, some packages might reserve manova for more complicated models than the multivariate T situations $\endgroup$
    – Glen_b
    Feb 6, 2014 at 9:38
  • 2
    $\begingroup$ See here. Many programs include that under manova, others do it separately $\endgroup$
    – Glen_b
    Feb 6, 2014 at 9:45
  • $\begingroup$ See stats.stackexchange.com/questions/190156/… for an alternative $\endgroup$ Nov 17, 2022 at 16:59

1 Answer 1

2
$\begingroup$

@Glen_b's short comment above essentially provides a full answer to your question, and it is difficult to expand it further. MANOVA is exactly what you need, and it will give you only one single p-value, precisely as you want. By the way, I have recently written up an explanation of what MANOVA does and how it is related to linear discriminant analysis, providing some illustrations. If you want an intuition behind how MANOVA works, you might be interested to take a look.

$\endgroup$

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.