Calculating correlation between dummy variables using pwcorr

[Guest]

Hi,

fist of all I was wondering whether pwcorr is okay to use when I would like to calculate the correlation between two dummy variables? Further I am not quite sure whether my type of interpreting the results is correct. Suppose we have two dummy variables. one is age, which takes on the values 1 = Female 0 = Male and the variable Riddle with 1 = Solves riddle and 0 = Does not solve riddle. Suppose I run pwcorr and got a significant (on the 5% level) correlation coefficient of -0.15. Do I have to interpret the result as a probability? In other words: The probability that a women solves the riddle is 15% lower than the average probability of men solving the riddle?

Thanks!

[Carlo Lazzaro]

Guest:
provide that -pwcorr- can be ok for your research purposes, your interpretation of its results is not correct, as correlation is simply another way to express covariance (that has nothing to do with probability).
A basic text on statistics can be helpful on this respect.
That said, you're possibly looking for -logistic- or -logit-:

logistic riddle i.gender

[Nick Cox]

This question is an echo of others posted elsewhere:

https://www.reddit.com/r/stata/comme...mmy_variables/

https://stats.stackexchange.com/ques...ical-variables

You got mixed up between gender and age, which in my experience are different variables.

If you are fteddy91 on Reddit, please note our longstanding advice to tell us about cross-posting elsewhere. https://www.statalist.org/forums/help#crossposting

Even if you aren't fteddy91 the question seems disingenuously phrased as a way to hide a class assignment. See also our policy on assignments. https://www.statalist.org/forums/help#adviceextras #4

If correlation is a probability you would be hard put to it to interpret negative correlations. In short, it isn't.