03-02-2017 12:19 PM - last edited on 03-03-2017 08:49 AM by ChrisHemedinger
I have a split-plot design: main-plot is fertilizer (Fert), sub-plot is the presence or absence of an N-fixing species (Nfix). We are testing the flux of gases from the soil. I have used PROC MIXED to analyze this design and that worked fine. Additionally, we included a Repeated statemnet due to measuring the same plots 4 times.
The complications arise when introducing what i believe is a hierarchical twist. In the Nfix sub-plot, i sampled by tree species: in both the Y and N levels of Nfix 2 trees were selected for sampling, with I and nI in the Y treatment and n1 and n2 in the N treatment.
I used the following code and the results do not return any value for Nfix or any interactions including Nfix. Additionally, all LSMeans values returned are P<0.00001, which absolutely cannot be true. PROC MIXED correrctly identifies 3 DF for the Fert level, but incorrectly assigns 497 DF for everything else.
So, how can i write a code that will correctly identify the DF in this split-plot analysis that includes the analysis by tree type in the sub-plot?
proc mixed; Class Block Fert Nfix Tree Date PlotID; Model logCH4= Fert Nfix Tree Fert*Nfix Date Date*Fert Date*Nfix; Random Block Fert*Block; Repeated Date / subject=plotID Type=CS; lsmeans Fert Nfix Nfix*Fert Date Date*Fert Date*Nfix/pdiff adjust=tukey;