1
$\begingroup$

I have conducted a series of tests on a single, dichotomic variable (presence/absence of a personality disorder in a neurological illness) in order to compare several things (i.e. presence of anxiety or depression, disease duration, gender, etc) between subjects who show the personality disorder and subject who don't. Since some of these variables are continuous and other not, i have used different statistical tests, using ChiSq for three dichothomic variables with presence/absence of something (anxiety, depression, gender, disease form) and mann whitney test for continuous variables (disease duration, disability level, MSFC). Now, is it correct to adjust the significance level for the number of comparisons (e.g. bonferroni), regardless of the kind of test being used, or do i have to choose a specific strategy for each test?

$\endgroup$

1 Answer 1

3
$\begingroup$

If the underlying personality disorder is a matter of degree (as many are) it is incorrect to analyze it as a dichotomous variable. To get at your question, though, it may be better to do a test of the global null hypothesis that none of the factors is associated with the personality disorder, by using a binary logistic model containing all factors as independent varibles, and computing the global chi-square test (the likelihood ratio test is the preferred one). This contains a perfect multiplicity adjustment. If that is "significant" you can dig deeper into partial associations.

$\endgroup$

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.