1
$\begingroup$

[first question, apologies any ambiguities/violations of question conventions]

Within a dataset, I have a continuous variable of interest (X), and I want to test for differences in variance of that variable between subgroups of interest. For example: does the variance of X differ between males and females?

After reviewing literature (see Lim & Loh, 1996), I decided to use the bootstrap-version of the Brown-Forsythe modified Levene test.

My dataset combines data from several different studies with different methods, ascertainment strategies, etc., so I first tested for variance differences as a function of study. I found a significant effect of study.

On the basis of that finding, I would like to test for significant differences in X across subgroups while controlling for the effect of study. Is there a natural, commonly used way to do this within the context of the Levene test?

Thank you!

$\endgroup$
2
  • $\begingroup$ ANCOVA is a test of means, so your mention of that method confuses me after you had been discussing testing groups for different variances. $\endgroup$
    – Dave
    Jul 11 '20 at 19:06
  • $\begingroup$ @Dave apologies, that was irrelevant to my question and so I removed. $\endgroup$ Jul 12 '20 at 13:13
0
$\begingroup$

In meta-analysis for variance Senior et al. discuss making "a statistical correction for differences in variance between groups (e.g. by using a standardized mean difference)" and cite Nakagawa et al. for that purpose. Senior et al. also present weighting considerations and those may be relevant to your question. For example, if one study presents data only for males, and another only for females, standardized mean difference would not be useful for comparing those studies. For the purpose of adjusting for weighting Senior et al. suggest a second paper by Nakagawa et al., which paper appears to offer adjustment of subpopulation mean value reference standards (e.g., between male and female) as well as overall meta-anaytic variance adjustment. This may be what you need.

$\endgroup$
2
  • $\begingroup$ this is very helpful! thank you. one small thing--the third hyperlink in your answer appears to be broken. could you re-post that link? $\endgroup$ Jul 13 '20 at 16:59
  • $\begingroup$ Fixed the link. Also, if this is useful, you should accept it as an answer and up vote it. $\endgroup$
    – Carl
    Jul 14 '20 at 2:26

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.