1
$\begingroup$

I have a series of continuous measurements that are not normally distributed (multivariate Shapiro-Wilk test p-value = 0.003818, Anderson-Darling test p-value = 0.0178 or lower for each variable) that are then stratified by several other ordinal data (such as: death Y/N, sex M/F, fever Y/N).

I understand that multivariate analysis can be done only with normally distributed data. The question is: are there not-normal alternatives? As there is the Mann–Whitney U test for the t-test, is there a non-parametric MANOVA?

Alternatively, what would be the procedure to analyze not normally distributed data?

$\endgroup$
1
  • 1
    $\begingroup$ How to do it depends on what hypotheses you want to test, but "multivariate analysis can be done only with normally distributed data" is definitely wrong. $\endgroup$
    – J.G.
    May 9 at 17:53

1 Answer 1

0
$\begingroup$

The Kruskal-Wallis test can be used. It is the equivalent of MANOVA in non-parametric tests.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.