topic Re: how can I test homogeneity of variances with more than 1 factor? in Statistical Procedures

how can I test homogeneity of variances with more than 1 factor?

palolix — Thu, 17 Jan 2013 07:59:27 GMT

How can I test homogeneity of variances to my ANOVA model when having more than 1 factor, using Levene for example?

Thank you very much!

Caroline

Re: how can I test homogeneity of variances with more than 1 factor?

SteveDenham — Thu, 17 Jan 2013 12:15:35 GMT

One way is to test a means model in PROC GLM, something like:

proc glm data=yourdata;

class factor1 factor2 factor3;

model response=factor1*factor2*factor3;

means factor1*factor2*factor3/hovtest=bf;

run;

This does Brown and Forsythe's version of Levene's test--testing at the finest granularity in your data.

Another possibility is in PROC GLIMMIX. See example 41.9 Testing Equality of Covariance and Correlation Matrices in the documentation. The likelihood ratio test is closer to a Bartlett's test than Levene's.

Steve Denham

Re: how can I test homogeneity of variances with more than 1 factor?

palolix — Thu, 17 Jan 2013 20:15:29 GMT

Thank you very much for your help Steve!!

Caroline