I would like to conduct a meta-analysis on the association of a continuous and a dichotomous variable. In order to compare the result with previous meta-analysis I woud like to use the point-biserial correlation as effect size. The primary studies provide different information. There are the following cases:
- a) correlation coefficient, sample size, sample size per group
- b) mean, sd and sample size for each group
- c) mean and sample size for each group, sd for continuous variable.
- d) 2 X 2 table
What would be the best way to compute the effect size and variance for each case? Is it recommended to use a Fisher´s r to z transformation and is there an easy way (within the metafor package) to do this with the point-biserial correlation?
I would be very grateful for any recommendations! Thank you very much.