Dummies in a correlation matrix

[Jul!a]

Hi everyone,

I am working on a multiple moderate regression. And just wanted to add the correlation matrix as well. Now I got the responds that I cannot state the mean and standard deviation of dummy variables. How does it work? Do I just leave the mean out and the standard deviation. Taking gender for example, or do I have to show the correlation not by gender but with male and female?

Thanks a lot!

 

[maartenbuis]

As long as you code your indicator variable with the values 0 and 1, then the mean is definately useful: it is the proportion of 1s. So, if you have a variable female coded 0=male 1=female, than the mean of the variable female is the proporiton of females in your dataset.

The sample variance of a binary variable is just a trivial transformation of the the mean: \( \frac{N}{N-1} \mathrm{mean} (1-\mathrm{mean}) \). So once you reported the mean of a binary variable there is no added value in added the variance of the standard deviation.

 

[Jul!a]

Thanks a lot! that is what I thought as well. So it is done that way in research papers? and then I leave the standard deviation out?

 

[maartenbuis]

Conventions differ by (sub-(sub-(sub-(...))))discipline. So, I would just look at the journal to which you want to submit your paper and see what others have reported in recent issues of that journal.

 

[Jul!a]

thank you very much!