www.cmm.bristol.ac.uk/forum

Hi,
I would be grateful if you could advise as I am struggling to get Realcom-Impute to start with even a most basic MOI and imputation model. I have a two-level model of interest with one explanatory variable, which has missing data. As a first step, I am trying to do an imputation model with the one response variable and cons as the only auxiliary variable. I encounter the following error (image in attachment): wrong parameters(AUXI "IDClass" DataGridViewTExtBoxCell {ColumnIndex=0, RowIndex=0}"}. ClassID works fine in MlwiN in my model of interest. (I have also tried adding extra auxiliary variables and extra variables to the MoI but to no avail).
Many thanks for any suggestions,
Lorraine

This looks like a bug in recent versions of MLwiN. I will look into it as soon as I can, however in the mean time if you request and install version 2.36 of MLwiN from https://www.cmm.bris.ac.uk/clients/softwaredownload/ then it should work there. Alternatively you could create the files required for Realcom-Impute directly using the commands documented in Goldstein H. (2009) REALCOM-Impute: Multiple Imputation using MLwiN, User Guide, linked to from http://www.bristol.ac.uk/cmm/software/r ... ation.html.

hi Chris,

Many thanks for this advice. I am certainly making more progress with the earlier version of MlwiN. However, I am still having problems at the stage of Realcom Impute writing values as I get an error message that it is 'unable to write file storedata: permission denied' (screenshot attached). I have specified in Estimation: impute a subfolder of the folder where my data are saved so I am not aware of any write restrictions there. I would appreciate your advice.
thanks
Lorraine

Hi,

Apologies as I see that in response to an earlier query, you have previously answered a very similar query to mine about write errors. I can see that the impvals are saved despite the error.
However, when I try to use the imputed values in MlwIn, I am getting various errors. The first relates to the precision of the data (even when I have specified that the response variable is categorical) and the second error relates to protected columns being overwritten. Screenshots are attached. I would be grateful for your help and apologies for so many questions.

Regards
Lorraine

Unless the variable in question is going to be used as a level identifier or is a categorical variable then it is safe to say no to the question regarding precision and the data will simply be rounded. If you say yes then new consecutive codes will be created, with the original values as category labels. Version of MLwiN later the 3 handle this more automatically, so it may be that you wouldn't get the message there.

The second error is due to the code for the models of interest writing into the columns usually reserved for saving the model results. I will investigate why this is appearing to the user (it may be that I have already fixed this in versions more recent than you are using), in the mean time if you enter the command: into the command interface prior to running the imputation then messages like this should be suppressed.

Yes, this is what I expected regarding rounding. Just as my MOI fails to run in MlwiN, I thought perhaps these are critical errors but clicking no is not a problem if the model then ran!

Regarding typing into command interface, this doesn't appear to be possible for me once the GUI has opened - is that unusual? There is a flashing horizontal line but nothing I type appears here. I can overlook the error messages, except that my model doesn't run with the imputations.

thanks for any suggestions.

Sorry - I meant the command interface for MLwiN (available via the Data Manipulation->Command Interface window). You are correct that you cannot type into the text window for Realcom-Impute and that this is there just for output. When you get the "overwrite protected column" message from MLwiN you need to click Yes, otherwise the results from the current model of interest will not be applied. If you were doing this then it would be useful if you could explain in what sense the model isn't running. Are you starting it via the Model->Imputation->Start Analysis menu option?

Hi,
thanks, yes I am running it via Model/Imputation/Start Analysis. What happens is that MlwiN indicates that it is still using the same subset of cases as before I did the imputation; i.e., it is using N cases with complete data on the variable which I have supposedly imputed so the model is still using 2119 of 2473 cases. Thank you.

The number of missing cases does not change because once the series of models of interest have been fitted the original data is brought back into the worksheet. You should however see blue estimates in the equation window that are different to those you would get if you fitted the model on the data with missing values. For a worked example see the Realcom-Imputation tutorial at http://www.bristol.ac.uk/media-library/ ... tation.pdf.

Sincere thanks for this. I thought the blue was the original MlwiN blue before getting the estimates in green. This clears up my confusion, many thanks again.