Using mi testransform on svy data with subpop command

[Jenjira Yahirun]

Hello! This is my first time posting so my apologies if this is not clear. I am trying to use the mi testtransform command in Stata 13.1 to test the equality of coefficients from interactions in a logistic regression model. I am using stratified and weighted data with the subpop command. However, the command is not producing any results and I am not sure if the syntax is correct.

Below is my syntax:

mi estimate, saving(miest, replace): svy, subpop(nosingcores): logistic momclose_4rv trimar_4r##race_ethr i.childimm i.female age_4 i.parent_4 i.educ_4 i.lfs_4 $mothertraits
mi test 2.trimar_4r#2.race_ethr 2.trimar_4r#4.race_ethr 2.trimar_4r#7.race_ethr
mi test 2.race_ethr 4.race_ethr 7.race_ethr

mi estimate (diff1:_b[1.trimar_4r#1.race_ethr]-_b[2.trimar_4r#1.race_ethr]) using miest
mi testtransform diff1

I have attached a snapshot of the log file from the testtransform portion, which look like this:


Transformations Average RVI = 0.0000
Largest FMI = 0.0000
Complete DF = 128
DF adjustment: Small sample DF: min = 0.00
avg = 0.00
Within VCE type: Linearized max = 0.00


diff1: _b[1.trimar_4r#1.race_ethr]-_b[2.trimar_4r#1.race_ethr]

------------------------------------------------------------------------------
momclose_4rv | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
diff1 | 0 (omitted)
------------------------------------------------------------------------------

. mi testtransform diff1
note: assuming equal fractions of missing information

diff1: _b[1.trimar_4r#1.race_ethr]-_b[2.trimar_4r#1.race_ethr]

( 1) diff1 = 0
Constraint 1 dropped

F( 0, .) = 0.00
Prob > F = .


Is there something that I am missing? Any help is much appreciated.

Thanks,
Jenjira

[Steve Samuels]

Welcome to Statalist, Jenjira! Please read the FAQ, especially FAQ 12, which asks that all code and results be inserted between CODE delimiters, described in the FAQ. The beginning delimiter is "[C O D E]" and the ending delimiter is "[/ C O D E]", with spaces removed.

[Jenjira Yahirun]

Hi Steve, thanks for the correction. Here is my code:

mi estimate, saving(miest, replace): svy, subpop(nosingcores): logistic momclose_4rv trimar_4r##race_ethr i.childimm i.female age_4 i.parent_4 i.educ_4 i.lfs_4 $mothertraits 
mi test 2.trimar_4r#2.race_ethr  2.trimar_4r#4.race_ethr 2.trimar_4r#7.race_ethr
mi test 2.race_ethr 4.race_ethr 7.race_ethr 
mi estimate (diff1:_b[1.trimar_4r#1.race_ethr]-_b[2.trimar_4r#1.race_ethr]) using miest
mi testtransform diff1

The test results shows up as blank table, similar to the problem reported here: http://www.stata.com/statalist/archi.../msg01303.html.
I think I have enough degrees of freedom as my sample size is about 4,655 respondents. In a separate analysis, I did not use the MI commands and instead extracted one multiply imputed dataset to work with and I did not have this problem using the -lincom- postestimation command to test for equality of coefficients.

Thanks,
Jenjira

[Steve Samuels]

Thanks for using CODE delimiters; they are most helpful in looking at output, which is what I asked for. I'm not expert in MI, so I hope that others will chime in. One thing I do see: the problem has nothing to do with mi testtransform. The problems show up in the results of the second mi estimate statement and, maybe, earlier. So, show everything.

Your lincom experience won't be relevant, unless you first svyset the extract, then run svy, subpop(): logistic on the extract. So show those results too.