Hello and thank to anyone who will help. I have a factorial experiment whith 2 factors, A and B (3*4) where there are 4 blocks. This gives a total of 48 experimental units. All 12 treatment combinations have been randomized over the 12 units, and this, in each of the 4 blocks.
My question is : which error term should I use to test the fixed effect (blocks are considered random). I suggest the following : effect d.f. Block 3 A 2 A*block 6 : error term to test factor A B 3 B*block 9 : error term to test factor B A*B 6 A*B*block 18 : error term to test interaction total 47 In some exemples I found, the 3 error terms are pooled into one, which is used to test all the fixed effect. What is the correct way to proceed. My design is in fact more complicated, including repeated measures, each block been measured three times. Thus I am using proc mixed of SAS to perform the analysis and I am specifying the random statement as follows : random block block*A block*B block*A*B which gives the d.f. presented before. I also include a repeated statement. Thanks, Joël Rivest ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at http://jse.stat.ncsu.edu/ =================================================================