Alright, this is what you can try:

Suppose there are 4 conditions, 5 subjects, and one covariate. In the following 3dRegAna script, you have a 20X7 design matrix.

(1) Add the covariate (after removing the mean) to end of each line (before the backslash).
(2) The 20 rows of the matrix is partitioned into 5 parts: the 1st 5 lines correspond to the 1st condition for the 5 subjects; the 2nd ...
(3) The 1st 3 columns are for the 4 conditions while the next 4 lines code for the 5 subjects.
(4) Make appropriate adjustment in your script for your experiment.
(5) With this script you can get pair-wise comparisons (and main effects). However, for something like condition 1 vs. 234 it is terribly challenging. It is really a headache for me to deal with such a situation, and that was why I started to develop 3dLME.

Good luck,
Gang


3dRegAna \
-rows 20 \
-cols 8 \
-xydata 1 0 0 1 0 0 0 \
-xydata 1 0 0 0 1 0 0 \
-xydata 1 0 0 0 0 1 0 \
-xydata 1 0 0 0 0 0 1 \
-xydata 1 0 0 -1 -1 -1 -1 \
-xydata 0 1 0 1 0 0 0 \
-xydata 0 1 0 0 1 0 0 \
-xydata 0 1 0 0 0 1 0 \
-xydata 0 1 0 0 0 0 1 \
-xydata 0 1 0 -1 -1 -1 -1 \
-xydata 0 0 1 1 0 0 0 \
-xydata 0 0 1 0 1 0 0 \
-xydata 0 0 1 0 0 1 0 \
-xydata 0 0 1 0 0 0 1 \
-xydata 0 0 1 -1 -1 -1 -1 \
-xydata 0 0 0 1 0 0 0 \
-xydata 0 0 0 0 1 0 0 \
-xydata 0 0 0 0 0 1 0 \
-xydata 0 0 0 0 0 0 1 \
-xydata 0 0 0 -1 -1 -1 -1 \
-model 1 2 3 : 0 5 6 7 8 \
-bucket 0 MyOutput