Close Home Forum Sign up / Log in

WHICH STATS TEST TO USE

L

Advice please

I'm trying to see if there's any kind of relationship between 2 sets of variables - set A contains 7 values (different kinds of job roles) and set B contains 5 values (different kinds of issues encountered in doing the job).

Correlation assumes the same number of items in set A and set B. Is there a test/technique that can be used in the scenario above?

K

Hiya, correlation doesn't assume the same number of values but the same number of individuals, i.e. everyone was asked the two questions. Though you should have at least around 10 values if you want to assume the variables are linear and use a Pearson correlation. The test you will use will also depend on the number of individuals you have.

D

I realise I'm coming to this thread a bit late, but I'm imagining that the job roles variable cannot be continuous or even assumed to have a latent continuous trait. It sounds nominal, infact both sets do. Therefore you can/should test for the existence of association using a chi-squared test and the strength of association should be given by something like Goodman-Kruskal lambda (which is probably the easiest to interpret nominal association measure for non- 2x2 contingency tables).

L

Hi Dovetailed

Thanks for the suggestions - you're not too late as I'm still pondering on this. Will follow up.


Cheers
LBaines

17071