Hello, I'm analyzing data collected from a Randomized Complete Block Design with missing observations, so I'm using Proc mixed (SAS 9.4). Within each block there is one fixed main plot factor (A) and one fixed subplot factor within each plot (B). I have been analyzing as a split-plot design with block as a random effect: proc mixed; class block A B; model Y = A B A*B; random block block*A; Is this correct? Also, I have another response where I took repeated measurements over time (time). It is also unbalanced. I have been using: proc mixed; class time block A B; model Y = time|A|B; random block block*A; repeated time/ subject= block*A*B type=un ddfm=KR; I've also been trying different covariance models with the repeated measures. With the Unstructured and Compound Symmetry covariance structure, the random statement shouldn't include the subject effect. Is the above random statement OK because they are different? Thanks.
... View more
