Hi thanks for your reply, Every year the whole experiment was like this: Rep1 Time3 time 5 ... time 0 Cv1cv2cv3 cv3cv2cv1 ... cv3cv1cv2 Rep2 Rep3 Rep4 Randomly assigned time and cv per rep My main plot =time (5) Sub-plot= Cv (3) Replicates = rep (4) Pe = measured in percentage Loc = site year I had a total of 60 measurements for my whole year. And this was done for 5 years Hope it's a little bit clear now. ... View more

Thank you for your input.   Indeed I post and old code, my mistake.   My experiment was done for 5 years at the same location. But not in the same area or plants. It was a randomized Split-plot design with the main plot =time and the sub-plot= cv and for each year it was measured the variable PE in each plot.   This is the code that I am using per year.   proc mixed data=xx order=data plots=all;  class rep cv time;  model pe=time cv time*cv/ddfm=satterth outp=ORP_data1 residual;  random rep time*rep/solution;  lsmeans cv time time*cv/ adjust=tukey;  run; proc plot data=ORP_data1; plot resid*pred; run; proc univariate data=ORP_data1 normal plot; var resid; run;   not sure the code to determine the homogeneity of variances without using Levene test with GLM.   proc glm data=xx; class time; model pe=time/ss3; means time /hovtest=levene (type=abs); run; proc glm data=xx; class cv; model pe=cv/ss3; means cv /hovtest=levene (type=abs); run;      Also, after the single analysis per year. what I need to know is if the values of PE (variable measure each year) can be combined over the years  using either proc mixed or proc glimmix   I was using this code but not sure if it's right   proc mixed data=xx covtest plots=all;   class rep loc cv time;   model pe=  time | cv/dfm=satterth outp=file2 residual;   random loc rep(loc) cv(loc) time(loc)/solution;   lsmeans cv time cv*time/diff; run;   I will really appreciate your help with these codes I am overwhelmed with the ammount of information thank you! ... View more