Here is a macro for a randomized complete block design (RCBD):

%macro rcbd(resp,trt,block,data); proc glm data=&data; /* randomized complete block design */ class &trt █ model &resp = &block &trt / ss3; lsmeans &trt / stderr pdiff; output out=diag p=pred r=resid; proc plot data=diag; /* diagnostic plots of pred vs. resid */ plot resid*pred=&trt; %mend rcbd;And here is a macro for a two-factor completely randomized design:

%macro twoway(resp,A,B,data); proc glm data=&data; /* two-way analysis of variance */ class &A &B; model &resp = &A | &B / ss3; lsmeans &A &B / stderr pdiff; lsmeans &A * &B / stderr pdiff out=lsm; output out=diag p=pred r=resid; proc plot data=lsm; /* interaction plots */ plot lsmean * &B = &A; proc plot data=diag; /* diagnostic plots of pred vs. resid */ plot resid * pred = &A; plot resid * pred = &B; %mend twoway;Once these macros are defined, they can be used as follows:

/* RCBD for two different responses */ %rcbd(yield,fert,location); %rcbd(temp,fert,location); /* Two-Way Analsys for different predictors */ %twoway(height,nitro,variety); %twoway(height,row,column);

/* include previously defined macros */ %include 'macro.sas'; /* RCBD for two different responses */ %rcbd(yield,fert,location); %rcbd(temp,fert,location); /* Two-Way Analsys for different predictors */ %twoway(height,nitro,variety); %twoway(height,row,column);

options ls=80 ps=40 mprint;This can be very handy to verify that macros are doing what they were designed to do.

Last modified: Thu Jun 8 17:37:09 1995 by Brian Yandell yandell@stat.wisc.edu