Missing Data/Imputation Discussion > Running Regressions with Imputed Data
Using what software? What output did you get?
Jeremy Taylor
Hi Jeremy!
I am curious about this as well. Also, I will be using a Sobel test of mediation. How do I get the pool results for this test using SPSS v. 20??
Many, many, many thanks again!!!
L
L (and Jeff),
I can't be sure about why IBM makes the decisions they do about the programming of SPSS, but my guess for why you couldn't get coefficients in that instance is because when using stepwise regression on multiple imputed datasets, we can't be sure that the same variables will be entered into the model in the same order (or at all), so it can't produce pooled results when the models aren't equivalent in each imputation.
As for the sobel, I think the best way to handle that would be to obtain a pooled estimate from the MI and then use the pooled estimates to just run one sobel test (just use an online calculator that can be found in various places, including at the link below).
CLICK HERE FOR WEBSITE THAT HAS SOBEL CALCULATOR
I hope that helps!
Jeremy Taylor
So I ran MI with my dataset, and I ran a stepwise forward multiple regression. The output gives a pooled result, but does not give a beta coefficient or changes in r-squared. Where to go from here in trying to get those numbers? Thanks.