Page 1 of 1

Gender control variable: case unequal group sizes

Posted: Wed Jun 05, 2019 2:48 am
by nshevchu
Hello,

In my analysis, I used gender as a control variable and I got a statistically significant result. The problem is that in my sample male to female ration is about 1:5 which does not reflect the ration of the users of an applications, i.e. respondents of the survey. Considering this, can I still claim that the statistically significant result obtained from the control variable is meaningful?

Thanks all for any advice on this topic.

P.S. among all existing topics, I have not found the one that answers this question. I apologize if I overlooked the same existing topic - a reference to it will be much appreciated.

Re: Gender control variable: case unequal group sizes

Posted: Thu Jun 06, 2019 8:23 am
by jmbecker
Having a significant control you know that male and female respond differently on your DV. Statistically, there is no problem in having more male or female respondents (i.e., unequal group sizes on the control). Conceptually, however, I would find it problematic that you samples so much more of the one group if you find that they behave differently. To get more precise estimates you may want to sample more of the currently underrepresented group.

Re: Gender control variable: case unequal group sizes

Posted: Fri Jun 07, 2019 9:55 pm
by nshevchu
Dear Dr. Becker,
Thank you so much for the response!