www.cmm.bristol.ac.uk/forum

Hello,

My MOI is a 2 level model. I have two variables in the MOI with missing values, and both of these are measured at level 2. Is it OK to run an imputation model using RealCom with the level 2 variables as the responses, and include all variables from the MOI (other than the 2 variables with the missing values) as the explanatory variables? The variables from the MOI include both level 2 and level 1 variables. So my imputation model would include both level 2 and level 1 variables as predictors, even though the response variable is at level 2. Is that OK?

thanks,
Caroline

Yes - that should be fine. Let me know if there are any problems.
Harvey Goldstein

Thank you. I created aggregated versions of my level 1 variables and am now including those aggregated variables in my imputation model. I am following the "Using REALCOM Impute and Stata" to export Stata data into REALCOM and then pull it back into Stata. When I export the data from Stata to REALCOM using the realcomImpute command, should the dataset I am exporting be the original dataset (of patients clustered in physicians, so there dataset is unique on the patient identifier and there are multiple rows for the same physician) or should it be a dataset at the physician level (only 1 row per physician, since the level 1 variables are now aggregated variables at the physician level)?

thank you!

Should be original 2 level dataset
Harvey