SPSS Statistics

Logistic regression with multiply imputed data in SPSS 27

  • 1.  Logistic regression with multiply imputed data in SPSS 27

    Posted Wed April 14, 2021 04:10 PM
    Hello, 
    I searched the discussion posts to see if my question has been discussed and I did not find any relevant threads.

    I have used the multiple imputation features in SPSS 27 to create an imputed dataset. I did NOT have a problem with using the multiple imputation process in SPSS. Instead, my problem is with key statistics missing from output tables for analyses conducted on the multiply imputed dataset. 

    I ran a binary logistic regression model (with 5 predictor variables) on the multiply imputed dataset. Several key statistics that I would like to report are not in the output tables. First, there are no model summary statistics (e.g., Omnibus tests of model coefficients, -2 log likelihood, Nagelkerke R Square, Hosmer & Lemeshow test) for the pooled data. This information is provided for the original data and the imputed datasets, but there are no pooled estimates for these statistics. Also, the Wald statistic is not reported for the regression coefficients in the pooled table. 

    Does anyone know of a workaround for this? How can I compute these statistics for the pooled data? 

    Thank you for any advice or guidance you can share.

    ------------------------------
    Jennifer Cole
    ------------------------------