Hello everyone, I am doing a smiulation study on analysis of spatially correlated data. My response conditioned on random components has a Poisson distribution with different means for 5 different treatments. Also the response has a spherical spatial structure. Since the response has different means depending on treatments I use GROUP=trt option in the RANDOM statement. Then the expected variance of the response will have 5 different values each for each treatment because variance depends on mean. However, the expected range of the response is same which is 6 for all the treatments. So when I use PARMS statement to specify initials for covariance parameters, I should give 5 values for variance components and one value for the ranges of all the treatments. And I don't know how to do this in GLIMMIX. There might be other ways such as writing two RANDOM statements one for variance and one for range. Again, I don't know if that works with the TYPE=sp(sph). Here is more detail about my work. The Poisson means for 5 different treatments are 10, 10.8, 11.6, 12.4, and 13.2. So the initial values of variance components are reciprocals of the Poisson means. There are 5 plots received each treatment and 9 sub sampling points within each plot. To specify spatial structure among subsampling points, the variable pos is used. Y is my response. So the GLIMMIX code that I have been using is PROC GLIMMIX; CLASS trt plot pos; MODEL Y=trt/ DIST=Poisson DDFM=KR; RANDOM pos/ SUBJECT=intercept GROUP=trt TYPE= SP(SPH)(row col); PARMS (0.1 6 0.09259 6 0.08621 6 0.08064 6 0.07576 6); RUN: Notice that in the above PARMS statement, I have given reciprocal of Poisson mean followed by the range 6 for each treatment. When I use the above code I get different estiamtes for the range of response for different treatments. However, I want to get a single estiamte for the range of the response. I will appreciate if anyone could help me to figure this out in GLIMMIX. Thank you in advance. Sincerely, Nadee ... View more