twopm with weights

[Karen Arulsamy]

Hi there,

I'd like to run a two part model as follows but can't seem to get the weights to work. The use of poisson distribution for the GLM model was based on the modified park's test.

Any advice much appreciated.

twopm tothealthcarecost i.ace_cat_b i.GENDER i.age_cat i.educ i.K1 i.F2 i.region i.ETHNICITY if sample == 1 & missing_ace == 0 [aweight=weighttrim], firstpart(logit) secondpart(glm, family(poisson) link(log)) vce(robust)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(tothealthcarecost ace_cat_b) byte GENDER float(age_cat educ) byte(K1 F2) float region byte ETHNICITY float(sample missing_ace weighttrim)
      0 0 2 4 0 3 6 0 3 1 0   5.377788
1900000 0 2 1 1 5 4 1 2 1 0  .02180295
      0 0 1 1 1 1 4 1 1 1 0 .011189446
2374000 0 1 3 1 4 4 1 1 1 0 .021811957
 450000 0 1 1 0 5 2 1 1 1 0   .7457822
      0 0 2 4 1 1 7 1 2 1 0   .0649344
      0 0 1 3 0 1 4 1 1 1 0  1.4537774
2950000 0 2 1 1 5 5 1 1 1 0  .02180295
      0 0 2 2 0 1 7 0 5 1 0   5.775696
      0 0 1 4 1 1 7 1 4 1 0 .033324864
1450000 0 2 1 1 5 4 0 5 1 0  .05074032
      0 0 1 3 1 1 7 1 1 1 0 .021811957
1600000 0 1 3 0 1 4 1 1 1 0  1.4537774
      0 0 1 1 0 1 6 1 1 1 0   .7457822
      0 0 1 4 1 1 5 1 1 1 0 .033324864
      0 0 2 1 1 5 2 0 3 0 0          .
      0 0 2 1 0 5 6 0 2 1 0   3.381867
1900000 0 2 4 1 4 4 1 1 1 0   .0649344
      0 0 2 3 0 1 5 1 1 1 0   2.832725
      0 0 2 2 1 1 7 1 2 1 0  .05413437
      0 0 2 4 1 4 5 1 5 1 0   .0649344
      0 0 1 2 0 1 4 1 2 1 0  1.8516984
 450000 0 1 3 1 1 7 0 3 1 0  .05076127
 474000 0 2 2 1 1 6 1 1 1 0  .05413437
      0 0 2 4 1 1 1 1 4 1 0   .0649344
 750000 0 1 1 1 5 4 0 4 1 0  .02604031
1450000 0 1 1 0 5 1 0 1 1 0   1.735601
 308000 0 1 1 1 1 1 1 2 1 0 .011189446
 150000 0 2 1 1 5 4 1 1 1 0  .02180295
 450000 0 2 2 1 1 1 1 1 1 0  .05413437
      0 0 2 3 1 1 7 0 4 1 0  .09890965
      0 0 2 3 1 1 4 1 2 1 0  .04250114
      0 0 2 2 1 1 7 1 1 1 0  .05413437
2074000 0 2 2 1 1 7 1 1 1 0  .05413437
      0 0 1 3 0 1 4 1 1 1 0  1.4537774
      0 0 2 3 1 1 7 0 4 1 0  .09890965
      0 0 1 3 1 5 5 0 5 1 0  .05076127
      0 0 1 1 1 5 2 1 1 1 0 .011189446
      0 0 1 3 1 1 7 1 5 1 0 .021811957
 150000 0 2 2 1 5 7 1 5 1 0  .05413437
      0 0 1 2 1 1 6 1 1 1 0  .02778219
      0 0 1 2 0 1 3 1 1 0 0          .
      0 0 2 2 1 1 7 1 1 1 0  .05413437
      0 0 2 4 1 4 7 1 5 1 0   .0649344
      0 0 1 4 1 1 5 1 1 1 0 .033324864
      0 0 1 2 0 1 6 1 1 1 0  1.8516984
1450000 0 1 1 1 1 6 1 1 1 0 .011189446
      0 0 1 1 1 1 7 1 1 1 0 .011189446
      0 0 1 2 1 1 7 1 5 1 0  .02778219
      0 0 2 3 0 1 1 1 5 1 0   2.832725
      0 0 1 4 1 1 7 1 5 1 0 .033324864
      0 0 1 1 1 5 5 0 5 1 0  .02604031
      0 0 2 3 0 1 4 1 5 1 0   2.832725
      0 0 2 4 1 3 7 1 2 1 0   .0649344
1450000 0 1 1 0 5 4 1 2 0 0          .
      0 0 1 1 0 5 2 1 1 1 0   .7457822
      0 0 2 2 1 1 1 1 1 1 0  .05413437
 158000 0 2 2 1 5 7 1 5 1 0  .05413437
      0 0 1 1 0 1 4 1 1 1 0   .7457822
 450000 0 2 1 1 5 7 1 1 1 0  .02180295
      0 0 1 2 1 1 7 1 1 1 0  .02778219
3950000 0 2 1 1 5 3 0 5 1 0  .05074032
      0 0 2 3 0 1 4 0 4 1 0   5.564115
 150000 0 1 1 1 5 3 0 1 1 0  .02604031
      0 0 1 1 1 5 4 1 2 1 0 .011189446
      0 0 2 2 0 1 7 1 1 1 0   3.608084
1450000 0 2 3 0 1 5 1 1 1 0   2.832725
3350000 0 2 1 0 1 6 1 1 1 0   1.453177
      0 0 1 3 0 1 4 1 1 1 0  1.4537774
      0 0 1 3 1 1 7 1 1 1 0 .021811957
      0 0 2 2 0 1 5 1 1 1 0   3.608084
      0 0 1 1 0 1 2 0 3 1 0   1.735601
1900000 0 2 1 1 5 2 1 1 1 0  .02180295
 150000 0 2 2 1 1 7 1 1 1 0  .05413437
3950000 0 1 4 0 1 5 0 4 1 0   5.169037
      0 0 1 2 0 3 3 1 1 1 0  1.8516984
4124000 0 2 2 1 1 7 1 2 1 0  .05413437
      0 0 2 1 0 1 5 1 2 1 0   1.453177
      0 0 1 2 1 5 3 0 5 1 0  .06465532
      0 0 1 2 0 1 6 1 1 1 0  1.8516984
      0 0 1 1 0 5 4 1 1 1 0   .7457822
 316000 0 1 1 1 1 6 1 2 1 0 .011189446
      0 0 1 2 0 1 4 0 5 1 0   4.309315
      0 0 1 2 0 1 6 1 2 1 0  1.8516984
      0 0 2 3 1 1 7 0 4 1 0  .09890965
 900000 0 2 1 1 1 6 1 2 1 0  .02180295
      0 0 1 3 1 1 7 1 2 1 0 .021811957
 450000 0 2 2 1 1 7 1 1 1 0  .05413437
      0 0 1 4 1 3 7 1 1 1 0 .033324864
      0 0 2 1 1 1 2 0 5 1 0  .05074032
 450000 0 1 2 0 5 1 1 1 1 0  1.8516984
      0 0 1 1 1 1 7 1 2 1 0 .011189446
      0 0 2 5 1 5 5 1 5 1 0   5.353928
      0 0 2 3 0 1 4 1 2 1 0   2.832725
      0 0 1 2 0 1 4 1 1 1 0  1.8516984
      0 0 1 1 1 1 1 0 3 1 0  .02604031
 158000 0 2 2 1 1 7 1 1 1 0  .05413437
      0 0 1 3 1 1 5 1 1 1 0 .021811957
2240000 0 2 3 1 1 1 1 1 1 0  .04250114
 450000 0 1 1 0 1 4 1 1 1 0   .7457822
end

Many thanks
Karen

[Leonardo Guizzetti]

All I can say on this is that your model implies one part that uses logistic regression (logit) and -logit- does not support -aweight-s. You will need to consider a more appropriate choice of weights here.

[Richard Williams]

Also, you are trying to estimate a count model in the 2nd part. Note that the help for twopm says

twopm is designed to estimate models in which the positive outcome is continuous. It does not deal with discrete or count outcomes.

I think the help file is a little confusing, because it lists all the standard glm options, but only a few work correctly with twopm. Further, there is no error trapping when you do use an illegitimate option.

You may be lucky you got the aweights error, because otherwise you might not have realized the twopm command doesn't do what you want it to.

[Richard Williams]

Are you sure you need a count model though? Would plain old regress be ok? twopm also has examples like this:

Two-part model with probit and linear regression of log(depvar>0)
. twopm wage educ age married children, firstpart(probit) secondpart(regress, log)

[Richard Williams]

On the other hand, Bill Gould does have this famous blog post on using Poisson rather than regress… so now I’m wondering if the two pm command is ok after all, once you take care of the weights problem.

https://blog.stata.com/2011/08/22/us...tell-a-friend/