Download example analyses from this page, to get you started with your own problems.
realcomImpute and Stata
This example illustrates how:
- data can be exported to the REALCOM Impute package from Stata
- multi-level multiple imputation can be performed using REALCOM Impute
- the imputations can be imported back to Stata for analysis, using Stata 11's mi capabilities
A zip file containing the example Stata dataset and a document describing the above can be downloaded here. Please note that this zip file does not contain the Stata export/import commands themselves - to download these see the instructions on the realcomImpute and Stata page.
This example takes you through a Bayesian analysis of multilevel data with a partially observed covariate. There is an example sheet to guide you through, and winBUGS code for four models:
- full data
- 'observed' data (after making some observations mising)
- missing at random analysis (i.e. imputing the missing covariates under MAR), and
- not missing at random analysis (i.e.\ imputing the missing covariates under NMAR).
Sufficient detail is given to enable those who have not used WinBUGS before to work through the analyses. These files are available as a .zip archive (290Kb) here. To download WinBUGS, see the WinBUGS website.
The instructions for our multiple imputation macros for MLwiN contain an example of using multiple imputation in a multi-level setting. The macros can be downloaded from here.