04-28-2014 05:53 PM
1) Did I write this coding correctly? I'm running a CRD with subsamples. Fixed variables: Treatment and Rows(Treatment). Random variables Phenology.
2) I collected subsamples over three years but the years should not be pooled. How can I ask SAS to analyze only one year without altering my data set by hand?
3) How can I ask SAS to compare each row within each treatment and phenology but not between them? For example, I would like to compare 'Phenology Lvl 1, Trt Lvl 1, Row Lvl 1' vs. 'Phenology 1, Trt 1, Row 2' but I do not want to view 'Phenology 1, Trt 1, Row 1' vs. 'Phenology 2, Trt 2, Row 1.'
PROC MIXED DATA = Abundance3Year METHOD = TYPE3;
CLASS TREATMENT RowOrder PhenologyRef;
MODEL PairsRecovdRate = RowOrder(Treatment);
RANDOM PhenologyRef PhenologyRef*Treatment;
LSMEANS RowOrder(Treatment) / PDIFF;
Thanks for your help.
04-29-2014 08:57 AM
1. Ought to work. I'm more used to the usual notation of A*B, rather than A(B) for the model and lsmeans statements, but they should work. I am curious why you would use type3 analysis rather than the default REML with this design, though.
2. A BY statement would give separate analyses by year. Be sure the data are sorted appropriately.
3. But phenology is a random effect, so if you want to do this, you will have to construct ESTIMATE statements to compare the BLUPs at each phenology. This could become tedious.