Stata2Mplus Error: variables have a variance greater than the maximum allowed of 1000000.*

[Yawo Kokuvi]

Hello,

I am trying to convert a dataset from STATA to Mplus. I used the stata2mplus program and converted the data. However, when I read it into Mplus, I get an error: One or more variables have a variance greater than the maximum allowed of 1000000.

The variables that are generating this error are those that have unique IDs and contain system weights, as in the following examples.

nhispid = 20000548840101
nhishid = 2000054884
serial = 32851
perwgt = 2735
sampwgt = 6881


one of the responses to my earlier posts suggested that Mplus usually have some difficulty dealing with long IDs. Could this be the case here?

I've also posted same question to the Mplus discussion list, but thought I will do same here, in case there is something I can do to change my input file generated by STATA.

Thanks again, cY.

[Sergiy Radyakin]

The shortest answer is probably create a new set of IDs: 1,2,3,4,5 and drop the original IDs.
Then convert to Mplus again and see what happens.

[Yawo Kokuvi]

Thanks very much, Sergly. I agree, a new set of IDs would be in the order.

What should I do about the weight variables, particularly, perwgt and sampwgt, which are really critical for taking account of my complex sampling structure. Wouldn't I be messing with my sample weights if I replace those with unique IDs?

thanks - cY

[wbuchanan]

You probably could handle it by specifying the necessary arguments to the usevariables parameters in the Mplus syntax and/or specifying those values directly as weights. There are similar parameters in Mplus to identify ID variables as well.

[Sergiy Radyakin]

Yawo,

if some of your observations are 1,000,000 more times more influential than others, you probably will not miss much if you drop the observations with the smallest weights. On the other hand it is a reason to double check where the data is coming from, as it is not common at least in the world of the household surveys.

Best, Sergiy

[Yawo Kokuvi]

Thanks very much, folks. I was able to resolve the issue.

it turned out that there were a number of weight (household weight, annual weight) and ID (household, family, etc) variables in the dataset that I do not need. Upon carefully reading of the NHIS technical manuals, I deleted / removed all those I do not need, with the exception of "sampleweight", and used the IDVARIABLE ooption to specify the IDs (in this case, nhispid). This helped resolved the issue.

I can now proceed with further work on the dataset.

thanks again, and happy easter to you all !!

rgds, cY.