2

I ran a linear mixed model (LMM) for my nested data and would like to check for the homogeneity of variance assumption.

As there is no a button for checking the homogeneity of variance at the LMM menu, I am wondering can I used the residuals generated from the model and ran a levene test of variance?

However, the residuals generated got negative value and seemed that levene test of variance only allow positive value?

Can I transform the residuals in order to conduct a levene test of variance, or should I transform the dependent variable?

I am really confuse and thank you so much for all the help.

Kam
  • 85
  • 2
  • 9
  • Levene and Brown-Forsythe tests you will find in Explore procedure (see there button Plots). – ttnphns Oct 02 '14 at 04:57
  • I found levene in Explore, but I have negative residuals, should I covert it to absolute value before carrying out the test? If the result is significant (p=0.47) but pretty close to 0.05, can I just accept homogeneity of variance, or should I carry out transformation? And I am confuse whether I should transform the residuals or the raw data? Thank you so much! – Kam Oct 02 '14 at 06:11

1 Answers1

2

There is no need to conduct a formal test. Just inspect the plots of residuals vs fitted values and perhaps an autocorrelation plot, and make an assessment based on those. A histogram of residuals and a QQ plot are also useful for assessing normality.

Robert Long
  • 53,316
  • 10
  • 84
  • 148