I know how to create a composite mean for subscales if you have raw data. My problem is that I’m doing a meta-analysis in which there is no raw data because I’m using means and standard deviations listed in articles. A lot of articles use a certain measure which has 5 subscales. Some of them post a composite mean and standard deviation which I can use to create my effect size. However, others DON’T post a composite mean and standard deviation – only means and sds for the subscales – I need a composite mean and sd to use along with the dv mean and sd to calculate my effect size. I’ve even tried to make sure that the data to calculate the effect size isn’t retrievable in some other way (e.g., F, t, already listed as some effect size). I have found equations that one can use if effect sizes for scales are available, but all of them are for combining different dependent variables where one calculates intercorrelations and does all kinds of stuff to end up with one effect size. I don’t know if this means correlations between the dv and the iv or just correlations between the dvs. If Rosenthal and Rubin’s method for creating a composite effect size for mutliple dvs could even work when used on iv’s, I don’t know what to do with that score once I get it (like if I make a composite self-compassion effect size, how do I use it to create an effect size between self-compassion and depression?)
Is it possible to create a composite mean and standard deviation from subscales if you do not have raw data?