SAS PROC MIXED

Kaz's SAS, HLM, and Rasch Model

SAS PROC MIXED

Multi-purpose PROC MIXED

People use PROC MIXED to do multilevel modeling, but off course we can use it for other simple purposes, such as to get group means.

Use it without a random statement, so the model is the same as OLS regression models. The advantage is that PROC MIXED allows you to use CLASS statement to specify what predictors should be treated as categorical variables (PROC REG doesn't allow CLASS statement.).

How to compare group means and do statistical tests on them--using LSMEAS statement. Advantage: Gives you statistcal testing on every possible group combination. Disadvantage (I am not exactly sure if DF is correctly calculated for statistical testing. Still the procedure is good for exploratory purposes.)

How to creage a graph that shows group values--based on coefficients obtained by PROC MIXED.

HOW TO GO FROM THIS (COEFFICIENT TABLE)

Obs Effect PRODUCT Estimate StdErr DF tValue Probt

1 Intercept 900.80 58.4495 15 15.41 <.0001

2 PREDICT 0.001092 0.008870 12E3 0.12 0.9020

3 PRODUCT BED -42.3210 14.3708 45 -2.94 0.0051

4 PRODUCT CHAIR -42.2655 14.3697 45 -2.94 0.0051

5 PRODUCT DESK -39.2746 14.3695 45 -2.73 0.0089

6 PRODUCT SOFA 0 . . . .

TO THIS (GRAPH THAT COMPARES GROUPS)

Go to Standard Usages of PROC MIXED

For information inquiry (AT) estat.us