Multilevel Point-Biserial Correlation?

[emstruong]

Hello,

Thank you for the great package.

I was just wondering whether it was possible to do multilevel point-biserial correlations? Specifically, my within-subject variable is one dichotomous (yes/no) and one continuous.

I think it should because of the relationship with Pearson's correlation, but I was also hoping to see whether there were any resources for how this is done in the correlation package?

There wasn't much of an explanation in the JOSS package...

Thanks

[mattansb]

I think this will technically work in the package, but it will no longer be a true point-biserial correlation since the binary variable will be dummy coded and then person-mean centered, resulting in a proportion deviation scale, such that for each person with a proportion $p$, 0 become $-p$ (because of $0-p$) and 1s become $1-p$.

[emstruong]

I see... Do you have any recommendations for something I can read for how your package computes multilevel correlations? I couldn't find much information based on the JOSS article and wasn't sure where to look next.