topic Re: mixed proc in Statistical Procedures

mixed proc

always-good — Mon, 20 Jan 2025 16:00:52 GMT

Hello everyone, When modeling with proc mixed in SAS studio, in the Random effects builder, what is the difference when the variable in the model effects, is in the intercept box, subject effect (covariance structures), or group effect, please?

Re: mixed proc

StatsMan — Fri, 07 Feb 2025 20:42:58 GMT

The Random effect builder in SAS Studio does a lot! The best advice is to look at how the code changes as you add effects in different ways. The subject effect crosses the subject effect with all other random effects on that RANDOM statement. It will appear as the SUBJECT= effect on the RANDOM statement. The group effect causes the RANDOM statement to fit the random covariance parameters on this statement separately for each level of the group effect. It will appear as the GROUP= effect on the RANDOM statement.

Re: mixed proc

always-good — Fri, 07 Feb 2025 23:22:07 GMT

Please explain how to interpret the output for a random effect. I mean, there is no probability of the effect like in the fixed effect, but there are covariance parameter estimates. How can I know if there is a significance? Thank you!

Re: mixed proc

always-good — Fri, 07 Feb 2025 23:24:04 GMT

There are lower and upper in the table

Re: mixed proc

Ksharp — Sat, 08 Feb 2025 06:21:15 GMT

" how to interpret the output for a random effect."

Just interpret random effect as fixed effect.

"there is no probability of the effect like in the fixed effect,"

You can get estimated coefficient with SOLUTION= option.

"How can I know if there is a significance? "

You could use option COVTEST to check whether the random effect is significant or not.

proc mixed data=sashelp.class covtest; /*option covtest check random effect is significant or not*/
class age;
model height=weight/solution ; /*option solution= get estimated coefficient of fixed effect*/
random int/subject=age solution; /*option solution= get estimated coefficient of  random effect */
run;

Here random effect AGE got P value 0.2017 , which means AGE is NOT significant .You can use only Fixed effect in model.