Please could anyone guide me on how to fit spatially correlated autoregressive correlation structure in row and column directions for RCBD design laid out in row-column two dimensional layout using proc mixed of SAS? I tried the script below but resulted to error messages
proc mixed;
class rep genotype row column;
model yield=genotype;
random row col;
repeated /type=AR(1)(row column);
run;
quit;
I have problems with the repeated statement where error message is coming from.
Thanks
I have never had any luck using both a Random and Repeated statement. I would use one or the other. If you have a two dimensional AR structure you will likely be better served with the spacial correlation structures. Putting the dimension variables in hte Random statement is not correct. The AR(1) structure specification you have does not allow the c-list variables. There are a number of the spacial structures that are AR based. You may also need a subject= specification.
You may want to try the methods implemented in the geoR, or R-INLA packages for R. A nice, simplified interface is provided in the geostatsp package (though the INLA models are not implemented yet). You can fit anisotropic models with these software.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis