Nonparametric/Ranking Data Help!!

#1
Hi all,

I am a PHD student attempting to come to grips with an analysis of some recent data. I have a repeated measures design. 220 students were presented with 13 different problem scenarios and then had to indicate who they would seek help from, ranking a list of seven help providers (ie. using the numbers 1-7, each number only used once per scenario). Help providers included GP, school counsellor, teacher, friend, parent, chaplain and psychologist.

When I did the original analysis I performed lots of Friedman Tests (one for each of the 13 scenarios) and then follow up Willcoxon (using Bon adjusted p i.e. .05/13). This allowed me to show how rankings differed between scenarios. However, it would be useful to also compare help providers within the scenarios, i.e. Friends ranked before GP in one scenario, but in another scenario GP ranked before friend. I can do this with Wilcoxon (as I can just put any two blocks into a pair), however I'm not sure if that would mean I should put all 13x7 = 91 blocks into the same Friedman Test first? Is there some sort of non parametric MANOVA I should do? I've treid to read about OLR and PERMOVA but wanted to check if this seems suitable for me before I attempt to understand one of them (they look soooooo complicated :( ).

Hopefully this all makes sense and someone can offer something.

Many thanks,

Kip.