Wilcox or T-test with percentages in R

Question

[Bine]

Hello,

I was wondering if anyone could advice me on the following:

I have two groups (males and females) and different response rates for a treatment. Now I would like to know if the difference between the response rates is statistically significant.

Response Rate   

Female   Male
20%      50%

I never used percentages before to calculate the wilcox/t-test, how would I deal with it in this case?

Thanks a lot!

Answer 1

[michael.ante]

If you don't have any further parameters (age, smoker vs non-smoker, income, ....); I would use the underlying data and treat it as two binomial distributions. These you can test if they differ as described here .

If your n is not too big, I'd use a Fisher's test.

[Bine]

Thank you very much for your response. Depending on the sub-analysis I have around between 20-100 individuals in the males and females group. Would this be a small n?

[michael.ante]

That's a small n.

[Bine]

Thank you, just in case someone else is looking for this. From above code I modified below for my code:

vals <- matrix(c(3, 1, 3, 2),
               nrow = 2,
               dimnames = list(Group = c("Female", "Male"),
                               Outcome = c("Responder", "Non-Responder")))
fisher.test(vals)