I am analyzing some sensory data from three replicate frying studies followed by storage of fried food, using both the fit model platform and the full-factorial repeated measures add-in, for my data that is a 2 treatments x 2 storage times x N panelists . There are 3 replicate fry studies (nominal). I have tried both entering the fry replicate as a fixed factor or not entering it (entering it improves the model R2 but doesn't change the problem I'm seeing). The data are continuous, but since they were highly skewed, I did a Box-Cox transformation (the problem is the same regardless of whether I transform the data). The predicted values always seem to be a discrete number, almost like they are put into a category depending on the variables, resulting in the patterns shown above. I have a feeling this is a simple error, but I do not know where. Please help!