Page 2 of 3

Re: issue retrieving imputation

Posted: Mon Jan 05, 2015 5:32 pm
by bthierry
Hello -- We have the exact same bug here. Any time frame for the next MLWin release? Thanks!

B.Thierry

Re: issue retrieving imputation

Posted: Mon Jan 05, 2015 5:56 pm
by ChrisCharlton
The plan is some time later this month, assuming nothing major comes up while testing recent changes. If you have access to Stata then the -realcomImpute- command (http://missingdata.lshtm.ac.uk/index.ph ... Itemid=102) will work correctly with the files output from Realcom. Alternatively if you need a solution sooner then I can provide the older version if you send me an email.

Re: issue retrieving imputation

Posted: Fri Jan 23, 2015 9:39 pm
by ChrisCharlton
MLwiN v2.32 has now been released (update available from http://www.bristol.ac.uk/cmm/software/m ... rades.html), which should no longer have this bug.

Re: issue retrieving imputation

Posted: Fri Mar 27, 2015 2:34 am
by yan1202yan
Dear Chris,
I also have the exactly same problem, and the version I am using is v2.30. This version should have no problem, right? I am so confused.

Re: issue retrieving imputation

Posted: Fri Mar 27, 2015 3:06 am
by yan1202yan
Actually, when I select the file containing all the imputed datasets, nothing happened. Even no error appeared.

Re: issue retrieving imputation

Posted: Fri Mar 27, 2015 10:09 am
by ChrisCharlton
I just tested this with an example on version 2.30 and it worked for me. What might be confusing you is that after choosing the directory in the Imputation->Retrieve Imputation menu there is no confirmation that reading the file was successful. You should find that if you have your model set up that selecting the Imputation->Start Analysis option will run the models with the imputed data as expected.

Re: issue retrieving imputation

Posted: Fri Mar 27, 2015 10:29 am
by yan1202yan
Dear Chris,
Oh, thank you so much! I need to try again and will tell you the result.

Re: issue retrieving imputation

Posted: Mon Mar 30, 2015 9:43 am
by yan1202yan
ChrisCharlton wrote:I just tested this with an example on version 2.30 and it worked for me. What might be confusing you is that after choosing the directory in the Imputation->Retrieve Imputation menu there is no confirmation that reading the file was successful. You should find that if you have your model set up that selecting the Imputation->Start Analysis option will run the models with the imputed data as expected.
Thanks Chris. I appreciate your help a lot. After I selected start analysis, a dialog box about pasting data problem appeared. I attached the dialog box here.
In the previous step I have specified the type of each response. Do I still have the problem of data converting? and how can I convert the data now? Thank you for your help.

Re: issue retrieving imputation

Posted: Mon Mar 30, 2015 11:18 am
by ChrisCharlton
This dialogue box will often appear when importing data into MLwiN, and simply indicates that MLwiN is unable to retain the full precision of the data after loading it. If the variables in question are not used as identifiers it is therefore normally safe to just click Done. What is puzzling in your example is that it appears to be importing all the data that meets this criteria into c50, which might indicate a problem matching the data in the imputed files with those in your worksheet. Could you check that the variable names match between these files?

Re: issue retrieving imputation

Posted: Tue Mar 31, 2015 7:30 am
by yan1202yan
Thanks Chris for your quick reply. Well, I have 5 variables needed to be imputed. They are ordered category, un-ordered category, normal, un-ordered category, and ordered category, and the fifth one is level 2 response.
I check the 10 imputed datasets. All of them have 5 columns with the same sequence, representing those 5 variables, and without variable names. Additionally, in the Name dialogue box, c50 doesn't contain any data.
After I click Done for several times, this dialogue box doesn't appear, but the results of the regression doesn't change at all, keeping consistent with the original results without imputation.