How to test the assumptions for linear mixed effect model

[Jie111]

Hi,

I am using linear mixed effect via proc mixed statement.

Someone listed five assumptions:

• Within-group errors are independent with mean zero and variance σ2σ2.
• Within-group errors are independent of the random effects.
• The random effects are normally distributed with mean zero and covariance matrix ΨΨ.
• The random effects are independent in different groups.
• The covariance matrix does not depend on the group.

I am wondering how to test these assumptions in SAS. Is there any article that I could refer to?

Thanks in advance.

[StatsMan · Accepted solution]

The PLOTS=ALL option on the PROC MIXED or PROC GLIMMIX statement will give you lots of looks at the residuals in your model. Outputting either the residuals or the EBLUPS for the random effects to a data set and plotting with PROC SGPLOT will give you more looks. Switching to PROC GLIMMIX gives you the COVTEST statement, which allows you to test custom hypotheses about the covariance parameters in your model. See the doc for that statement here