0
$\begingroup$

I have a data set classifiable by gender (male and female) and genetic status (WT and mutant). The WT data is mostly consistent with tight standard deviations, whereas the mutant data is highly inconsistent with variable standard deviations. The n is equal for both sets. I am interested in the following - 1) running a statistical test of the standard deviations showing that the mutants are inconsistent with high variability (Levene's test?) 2) running an ANOVA (or ANOVA like) test for the parameters to see if the mutant and WT means are significantly different. How should I go about this?

Any help will be highly appreciated.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Tests of homogeneity of variance only test whether the population variances are exactly equal which, of course, they never are. ANOVA is very robust to violations of homogeneity of variance when you have equal n. The Welch test could be used but it is not necessary with equal n.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Thank you! I want to show the differences in standard deviations as a proof of how erratic the mutant mice are. Is there a test you would recommend for this? Is it incorrect to treat the standard deviations as a parameter and run an ANOVA on these SD values? $\endgroup$
    – swataus
    Commented Jun 2, 2017 at 22:43
  • $\begingroup$ If I understand you correctly, you have an sd for each subject. If so, you can do an ANOVA on them since the numbers don't remember where they came from. Of course, consider assumptions keeping in mind ANOVA is generally robust. $\endgroup$
    – David Lane
    Commented Jun 2, 2017 at 22:52
  • $\begingroup$ You are correct, I have an SD for each subject. I will go ahead and run ANOVAs on them. Thanks a lot! $\endgroup$
    – swataus
    Commented Jun 2, 2017 at 22:57

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.