direction of association between nominal and ordinal variables


I have a question about which test are appropriate to find direction of association between a nominal and an ordinal variable.

I have 2 kinds of participants (A and B), each of them answered a few questions with an answer on a scale from 0-7.
The sample size is quite small so I used a fisher exact test to find out if the responses were significantly different.

Now I would like to know, for the siginificant differences, in which direction the difference lies. For example do people from A choose lower numbers on the scale than people from B?

Is there any test that could give me the answer?

Thanks a lot!


Fortran must die
What do you mean by direction? Normally this means one variable drives differences in another, although from your comments I am not sure you mean it that way.

I think direction has to be part of your theory. For example in regression it is assumed that the IV cause the DV rather than the reverse because your theory says so. The regression model itself won't tell you that (I am not sure any model will tell if the direction goes both ways although maybe SEM does).

If you simply want to know if people on A chose different numbers than B (and what they chose) you can do a contingency table which shows what percent of A and B's choices were for each scale item (that is how many 0's as a percent of total, how many 1's, etc).