Multiple Imputation and Coefficient of Determination

[stodo53]

Hi all,

I have created 20 multiply imputed datasets, and have combined the estimates using proc mianalyze. I was wondering if anyone knows how to compute the average or "pooled" R^2 value from the 20 imputed datasets? 

Thanks!

[SAS_Rob]

One option might be to do as this page suggests, namely, treat the R-Square as you would a correlation coefficient.  This would allow you to compute a standard error using Fisher's Z transformation.  Once you do that, you could follow this example in the MIANALYZE documentation for combining correlation coefficients.

[stodo53]

I think that page uses STATA, is there anyway to do this with SAS?