02-18-2016 09:41 AM - edited 02-18-2016 09:44 AM
I am analyzing data from a randomized block design study. We have two independent variables (Treatment and Site), two covariate and one blocking factor (categorical). As I use a block design, I wonder what is a proper way to set up an ANCOVA model. I use SAS EG to do the analyses but I've told that mayneed to modified the code myself to include a blocking factor. Do I just add one more factor in the classification variable option in Linear Model? How can I make sure that SAS would treat the block variable separately from the two independent variables I have? Below is my code from SAS EG.
PROC GLM DATA=work1;
CLASS Treatment Site Block;
MODEL Tb_BVTV= Treatment Site Treatment*Site Block Covariate(1) Covariate (2)
LSMEANS Treatment Site Treatment*Site / PDIFF=ALL ADJUST=TUKEY ;
Any advice would be appreciated.
03-29-2016 08:42 AM
You may have received an answer already, but this is a classic case where a mixed model approach is superior to PROC GLM. See Littell et al.'s SAS for Mixed Models, 2nd ed. for examples. In your case, I would suggest
PROC MIXED DATA=work1; CLASS Treatment Site Block; MODEL Tb_BVTV= Treatment Site Treatment*Site Covariate(1) Covariate (2) / SOLUTION; RANDOM Block; LSMEANS Treatment Site Treatment*Site / DIFF=ALL ADJUST=TUKEY ; RUN;