Hi,
Would anyone be able to advise on the following please: I have used RealCom Impute to impute missing values at level 1 in my dataset and successfully created a model in MlwiN using the imputed values. I'm interested in the Level 2 residuals and I'm wondering if MlwiN can combine the estimates for the residuals across the different imputed datasets? If not, would I need to do the model with each of the imputed datasets, save the residuals and combine them myself using Rubin's rules? I'm currently using SPSS and MlwiN and I don't have access to Stata. any suggestions would be very gratefully received!
thanks
Lorraine
Lorraine Gilleece
Faculty of Education
Liverpool Hope University
Level 2 residuals after using multiple imputation
-
Harvey Goldstein
- Posts: 49
- Joined: Sun Sep 06, 2009 5:30 pm
Re: Level 2 residuals after using multiple imputation
Mlwin will not automatically combine residuals. You will need to do this manually. We will look into doing this.
Harvey Goldstein
Harvey Goldstein
-
GeorgeLeckie
- Site Admin
- Posts: 432
- Joined: Fri Apr 01, 2011 2:14 pm
Re: Level 2 residuals after using multiple imputation
Hi Lorraine,
I appreciate that you do not yet have access to Stata, however other users who do use Stata might want to consider using the user-written -realcomimpute- command in Stata to call the RealCOM-Impute software seamlessly from within Stata and to automatically generate the multiple imputed datasets. They can the use the user-written -runmlwin- command in Stata to fit the multilevel model of interest in MLwiN to each of these multiple imputed datasets. Users can then use the -mi- suite of Stata commands to both automate the process of fitting the model of interest to the multiple datasets and to automatically derive the combined parameter estimate and standard errors according to Rubin's rules. It would also be easy to extract the residuals from each -runmlwin- multilevel model, save them as new variables in a Stata dataset and then apply Rubin's rules manually to these quantities using standard Stata commands.
The realcomimpute command is documented at
http://missingdata.lshtm.ac.uk/index.ph ... Itemid=102
The runmlwin command is documented at
http://www.bristol.ac.uk/cmm/software/runmlwin/
Best wishes
George
I appreciate that you do not yet have access to Stata, however other users who do use Stata might want to consider using the user-written -realcomimpute- command in Stata to call the RealCOM-Impute software seamlessly from within Stata and to automatically generate the multiple imputed datasets. They can the use the user-written -runmlwin- command in Stata to fit the multilevel model of interest in MLwiN to each of these multiple imputed datasets. Users can then use the -mi- suite of Stata commands to both automate the process of fitting the model of interest to the multiple datasets and to automatically derive the combined parameter estimate and standard errors according to Rubin's rules. It would also be easy to extract the residuals from each -runmlwin- multilevel model, save them as new variables in a Stata dataset and then apply Rubin's rules manually to these quantities using standard Stata commands.
The realcomimpute command is documented at
http://missingdata.lshtm.ac.uk/index.ph ... Itemid=102
The runmlwin command is documented at
http://www.bristol.ac.uk/cmm/software/runmlwin/
Best wishes
George