Three -level Multiple Imputation

[helanidilk]

MCMC Manual- Chapter 18 describes a Multiple Imputation procedure for two-level continuous data. Here, the responses are treated as a lower level so the model becomes three-level. How this can be used for a three-level multiple imputation?

[billb]

Simply add an extra level i.e. to fit a 3 level multivariate model in MLwiN one simply asks for 4 levels with the additional lowest level being used to represent the responses within an individual. Hope that helps.
Bill.

[guobl]

HI Bill, regards to the Hungarian data imputation example shown in your MCMC manual, is there a code to unstack the imputed data stored in c31, c32, c33, c34, and c35 for each outcomes with MI dataset indicator and ID variables? sorry I am not familiar with MlwiN command. thanks. boliang

[ChrisCharlton]

If you generate the imputation via R2MLwiN (http://www.bristol.ac.uk/cmm/software/r2mlwin/) or runmlwin (http://www.bristol.ac.uk/cmm/software/runmlwin/) then this is automatically handled for you.

To do this directly within MLwiN you will need to use the SPLIt command (see the help file). The column indicating which response corresponds to each row already exists in the form of resp_indicator in the data, so to split the first imputation stored in c31 you just need to issue the following command (or if you prefer to use the graphical interface this corresponds to the Data Manipulation->Split column menu):

Code: Select all

split c31 'resp_indicator' c40-c45

You can then name the output columns (c40-c45) to indicate which response and imputation they each belong to:

Code: Select all

name c40 'mi1_es_core' c41 'mi1_biol_core' c42 'mi1_biol_r3' c43 'mi1_biol_r4'  c44 'mi1_phys_core' c45 'mi1_phys_r2'

You then just need to repeat this procedure for each of the other imputation columns (c32-c35), choosing new unused output columns each time.

[guobl]

Thanks Chris.boliang