Using Mibeta to Estimate R-Squared for a Model Estimated with Multiply Imputed Data

[Jason Benjamin]

Hi everyone, I recently installed the mibeta command in order to estimate R-squared for the pooled results of an OLS regression that was conducted with multiple imputed data. I originally performed the multiple imputation in SPSS and then converted it to a Stata/SE 13.1 file.

I then successfully entered the following command to ensure that the multiply data was mi set:

mi set flong, M = 5

There were originally 5 iterations for the multiple imputation that was performed in SPSS.

However, when I tried to perform the OLS regression using the mi beta command (Mibeta X Y), an error message appeared indicating that there were no imputations.

I'm wondering if the multiple imputation would have had to been originally performed in Stata for the mibeta command to work? Or is it fine if I converted the multiply imputed dataset from an SPSS file to a Stata file? If so, how can I use mibeta to estimate R-squared for the pooled results of an OLS regression model? It doesn't seem possible to do this in SPSS.

Any help would be greatly appreciated.

[William Lisowski]

If you look at help mi import, you will see that Stata has expectations about the format and content of multiply imputed data. So in general (with the exception of data for which mi import is designed) you need to do the imputation within Stata.

[Jason Benjamin]

If you look at help mi import, you will see that Stata has expectations about the format and content of multiply imputed data. So in general (with the exception of data for which mi import is designed) you need to do the imputation within Stata.

Thanks a lot, this is very helpful