2
$\begingroup$

I have a cell array containing the gender e.g.

GENDER = {'Male';'Female';'Male';'Male';'Male';'Female';'Female';'Female';'Male';'Male'};

and another array containing the respective groups e.g.

GROUPS = [1; 2; 1; 1; 3; 3; 1; 2; 2; 3];

I would like to find whether there are significant differences between the 3 groups in the number of male and number of female.

Or maybe, even better, the differences between the ration male/female in the 4 groups.

I think that the GENDER variable could be assumed Bernoulli-distributed so using the ANOVA should be statistically incorrect.

If instead of the GENDER I've to analyse the difference in eye colour what is the best approach?

EYE_COLOR = {'Blue';'Blue';'Brown';'Black';'Black';'Blue';'Brown';'Black';'Black';'Black'};

Can I assume the EYE_COLOUR a categorical distribution?

Thanks a lot

$\endgroup$
2
  • $\begingroup$ You're going to need to provide more context about your data and the question you are trying to answer to get a meaningful response. You should consider providing a sample of your data to help contextualize your question. $\endgroup$
    – David Marx
    Nov 18, 2013 at 16:50
  • $\begingroup$ I've edited the question...it's only an example..I know that with this data there are no significant differences... I hope that now is more clear $\endgroup$
    – Gabboshow
    Nov 18, 2013 at 16:53

1 Answer 1

1
$\begingroup$

Both variables are categorical without natural ordering. A $\chi^2$-test of independence (and its versions) is usually the test of choice in such situation. It tests the null hypothesis that the true proportions of females (and males) are equal across eye color versus the alternative that there are some differences.

(Of course, a non-significant result does not mean that the proportions are equal.)

If you compare the p value of the $\chi^2$-test with the F-test of the one-way ANOVA with 0-1 coded gender as response, you will get about the same. So your intuition was quite right!

Edit: The $\chi^2$-test is symmetric in its two variables. So you can as well check the null hypothesis that the color distribution of males equals the color distribution of females.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.