"Attempted to access MISSINGFULL(1,;)" error.

[Jamoo]

Hello,

I have exported a small set of 4 response variables and 2 explanatory variables to build a test imputation model.

I am using Stata 12 and exporting data to Realcom using Realcom-Impute using the following code:

Code: Select all

*export MI
ge cons= 1 //needed for realcomImpute model
sort ParticipantID
preserve
foreach w in childht1clean childht2clean childwt1clean childwt2clean {
	replace `w' = 99999 if `w' ==.
}
realcomImpute childht1clean childht2clean childwt1clean childwt2clean childage1 sex using temp.dat, numresponses(4) /*
        */ cons(cons) level2id(ParticipantID) replace
restore

The error message that I've received from Realcom after running with default settings (apart from the missing value which I specified as 99999) is below.

Screenshot 2015-04-28 22.00.29.png (165.57 KiB) Viewed 75388 times

I was also going to attach the .dat input files and the .dat model files which are the formats that Realcom uses but the Forum says that the .dat extension is invalid...

Happy to provide any further information that might be of help.

With best wishes

Jamie

[ChrisCharlton]

Could you please send the data file that you used as input for Realcom-Impute to Professor Harvey Goldstein, the author of Realcom (you can find his address here: http://www.bristol.ac.uk/cmm/team/), and he will take a look at what might be causing this? For future reference, I think that the forum may have accepted your file if you zipped it first (but I haven't tested).

[Jamoo]

Hi Chris and Harvey,

Harvey was very helpful in making some suggestions such as ensuring that all my observations had some data (which was an issue) and ensuring that I had Matlab stats toolkit installed (which I didn't..). However, these didn't resolve the issue.

I eventually 'solved' this in the sense that I realised that my data was still in wide format and not multilevel. When I reshaped to long format the model ran fine. Does the error message now make sense to you?

Best wishes

Jamie

[ChrisCharlton]

We think that this might have something to do with the fact that you have changed the default missing value. Did you remember to click the "Set value for missing to" button after changing this. If not then this might explain the message that you were seeing.

[Jamoo]

Thanks for continuing to look into this. Initially I did change the missing value. However, changing it to the default didn't resolve the issue so I ruled that out at the time.

The thing which seemed to fix it was that my data weren't actually multilevel initially (!) in that they were still in wide format and I'd given Realcom the data as if it had been reshaped long and outcomes were nested under participant id (my level 2 group variable).

Hope this helps.

Jamie

[eliotrasia]

Hi Chris and Harvey,

Harvey was very helpful in making some suggestions such as ensuring that all my observations had some data (which was an issue) and ensuring that I had Matlab stats toolkit installed (which I didn't..). However, these didn't resolve the issue.

I eventually 'solved' this in the sense that I realised that my data was still in wide format and not multilevel. When I reshaped to long format the model ran fine. Does the error message now make sense to you?

Best wishes

Jamie

I am using Stata 12 and exporting data to Realcom using Realcom-Impute using the following code: