nominal data analysis - hypothesis test

[Roseanna]

Hello!
Hopefully somebody will take pity and answer my question, it is driving me nuts and guess what? My supervisors seem to have no clue as well.

The study is a survey
My dependent variable is a scale of efficacy which will yield interval data.
My independent variable - yields only nominal responses to a question (which type of feeding method)(eg my survey results might be that of 400 women, 200 women used 260 feeding methods, and two of those methods were most frequent)

My burning question is - with two levels of measurement, is it possible to test the hypothesis that higher levels of efficacy is associated with less use of feeding methods.

Any help or guidance sincerely appreciated.
I am 3 years into this wretched course and can't quit now, but am sorely tempted.

[vinux]

I am not familiar with biology part.
I guess your dependent variable is continues and independent variable mainly has two values. ( I am bit confused with this part 200 women used 260 feeding methods )

In that case you use t-test or nonparametric comparison test like Mann-Whitney Test

[Roseanna]

Yes this the difficult part. Making a human topic fit a scientific model.

And yes, the dependent variable would be continuous.
My independent variable will be 'use of feeding methods', and all the participants will choose from 9 categories, and they might choose more than one. that's how I end up with possibly 200 women making 260 choices of category. (and 200 used nothing)

Now to sound like a real dummy is the mean figure 260 divided by 400?

If a t-test or Mann-Whitney is the answer for the right hypothesis test, that makes me very happy , because I have already written and then deleted many pages or correlation, and ANOVA.

Thanks for your help!

[spinningfeather]

If a t-test or Mann-Whitney is the answer for the right hypothesis test, that makes me very happy , because I have already written and then deleted many pages or correlation, and ANOVA.

If a t-test is appropriate then correlation, ANOVA, or regression, will give you the exact same results anyway (in terms of statistical significance).

[Roseanna]

Hi spinning feather,

the reason I had to change from correlations was because one of my supervisors is adamant that correlations cannot be used with nominal/categorical data to test the hypothesis. Is he very wrong?

If I can calculate a mean from categories, then I should be able to proceed (without being shotdown in flames at my confirmation seminar later?)

I just re-submitted everything yesterday proposing the t-test, just to make it easier on myself and get things moving. In my heart I wanted to continue with correlation to test the hypothesis.

Thanks for your reply,