05-27-2015 06:02 PM
I am working on sas code to test for treatment effects on soil pH through time.
I have 4 plots (reps), two of which are fertilized, and two that are not fertilized (Treat)
1/2 of each plot was plowed (plow)
I measured soil pH at four uneven time intervals (DOY), so end up with 32 data points.
Here is the code I developed - am I missing something? Thanks!
proc mixed data=soil; class treat plow reps doy;
model pH = treat|plow|doy / Lcomponents;
repeated doy / sub=treat type=cs;
05-28-2015 02:21 PM
I agree with Peter, the code formulation is perfect. I would hesitate on specifying a compound symmetry error structure, as it assumes equal correlations between days, no matter how far separated in time, and equal covariances at each time point. However, because of the small N for timepoints, and unequal spacing in time, this may be the only candidate that will be able to be fit.