How to Analyze: Comparing two proportions


I need advice on how to analyze a simple set of data. For simplicity's sake I'll make up the variables.

Question: A store keeps track of the proportion of males vs. females that visit its store every day. On Saturday, the proportion of men and women visiting is 0.5 and 0.5. On Sunday, it is 0.7 men and 0.3 women.

If this is pertinent, let's say that 400 visited on Saturday and 300 visited on Sunday, So that breaks down to, Saturday: 200 men and 200 women. On Sunday, 210 men and 90 women.

Is the difference between the two days significant?

What would be the appropriate test to use to analyze this?

There are at least two general ways to do this. One is chi-square. The other is logistic regression. The first asks about association. The latter about dependence. The first treats both variables equally; the latter treats one as dependent and the other as independent. The latter gives you an odds ratio (and a CI around it) that are fairly easy to interpret, the former gives you a chi-square value which is much less intuitive.

And they can give different answers.


Less is more. Stay pure. Stay poor.
Good point PeterFlom.

I will mention that with the former you can also get the odds ratio with an additional step, and also use the standardized chi-square residuals to better understand differences. Lastly, novices may find logistic regression a difficult procedure to perform if not acquaint with it.

Though, I really like logistic regression and appreciate the suggestion.