ok I have 2 groups with 20 people in them each.

I want to compare how many times they have mentioned topics x, y, z in a positive or negative way in interviews (X- X+ Y- Y+ etc)

So I wanted to do 6 t-tests (x, Y, Z) but now doing Mann Whitney cos of small sample.

However, group 1 generally mentions X, Y Z A LOT MORE both positively and negatively. So I want to do proportions rather than raw stats.

So participant 1 mentions X 20 times, of those mentions 5 were positive and 15 were negative so their score

X- = 25%
X+ = 75%

Then I want to use the mann whitney on the proportion data

IS THIS OK?? - need stats guru please :p

scrap that - I might have to use chi-square cross tabs? any ideas people????

Hi Sneaks, I don't know loads about stats but I think it's very difficult to do any tests on proportional data. the Mann whitney is a rank test so it might work best on the raw data...is there any way you could do a PCA? it's a really basic way of looking at how your data groups together based on several factors at once and can show you immediatly if there are any relationships. I haven't got my trusty stats book here though so I'll have a look tomorrow and get back to you if that's ok, I know it's annoying having to wait and see!

Thanks algequeen! I'm familiar with PCA - I only have 20 participants, I think PCA needs 100 per 10 questions for questionnaire analysis. These are qualitative interviews, so am counting codes, under predefined categories. I'm going to have a look at cross-tabs chi square tonight and see if it answers any questions!