Setting up an ordinal logistic regression for repeated measures

I have been looking for a non-parametric equivalent to a 3-way repeated measures ANOVA, and most of the sources I've read recommend using ordinal logistic regression.

My ANOVA model had 3 within-subject factors (time - 6 repeated measures; task - 3 tasks; 2 sides of the body) which gave 36 responses for each of the 9 subjects involved.

The matlab function mnrfit which performs OLR just takes a matrix of predictor values (in this case, the columns of which would be the categorical time, task and side vectors which form the full 36 response combinations e.g. first row would be t1, task1, right side etc. for the remaining rows) and the vector of corresponding responses.

How do I incorporate the 9 subjects into this model? Would it be appropriate to simply repeat the same matrix of predictor values (i.e. the 36 x 3 combinational matrix) 9 times with the relevant responses in the response vector?