Choice models

[Richmond Baye]

I have this example data that I have reshaped from wide to long. I want to use it to compare different choice models, specifically, the multinomial probit, mixed logit model, and random parameter model. The problem is that I keep getting the error "depvar choice must be a 0/1 variable indicating which alternatives are chosen." I would appreciate some assistance.[

* Example generated by -dataex-. For more info, type help dataex
clear
input double(cost indemnity diseaseuncertainty) str17 responseid long choice str7 alternative2
1.5               .75 0 "R_12c0WhpSacmBfND" 3 "choiceA"
1.5               .75 0 "R_12c0WhpSacmBfND" 2 "choiceB"
1.5               .75 0 "R_12c0WhpSacmBfND" 3 "choiceC"
  0               .75 1 "R_130tjeDa0SB0a7n" 2 "choiceA"
  0               .75 1 "R_130tjeDa0SB0a7n" 2 "choiceB"
  0               .75 1 "R_130tjeDa0SB0a7n" 4 "choiceC"
  3 .8999999761581421 0 "R_17ro9DXzTCsyb4d" 3 "choiceA"
  3 .8999999761581421 0 "R_17ro9DXzTCsyb4d" 4 "choiceB"
  3 .8999999761581421 0 "R_17ro9DXzTCsyb4d" 4 "choiceC"
  3 .8999999761581421 0 "R_1B19p1QowTgqSm5" 3 "choiceA"
  3 .8999999761581421 0 "R_1B19p1QowTgqSm5" 4 "choiceB"
  3 .8999999761581421 0 "R_1B19p1QowTgqSm5" 4 "choiceC"
1.5               .75 0 "R_1H7wPWuH1CbkDmD" 3 "choiceA"
1.5               .75 0 "R_1H7wPWuH1CbkDmD" 4 "choiceB"
1.5               .75 0 "R_1H7wPWuH1CbkDmD" 3 "choiceC"
  .                 . . "R_1Ino1LWi1nJIdUn" 3 "choiceA"
  .                 . . "R_1Ino1LWi1nJIdUn" 2 "choiceB"
  .                 . . "R_1Ino1LWi1nJIdUn" 4 "choiceC"
  3 .8999999761581421 0 "R_1JFmXqFfifSI8p2" 3 "choiceA"
  3 .8999999761581421 0 "R_1JFmXqFfifSI8p2" 4 "choiceB"
end
label values diseaseuncertainty diseaseuncertainty
label def diseaseuncertainty 0 "low", modify
label def diseaseuncertainty 1 "high", modify
label values choice choice
label def choice 2 "1", modify
label def choice 3 "2", modify
label def choice 4 "3", modify

]

[Devra Golbe]

It would be helpful if you show the command(s) that generated that error message.

[Richmond Baye]

sure! this is code.
"cmmprobit choice cost indemnity diseaseuncertainty"

[Devra Golbe]

Please see the FAQs about how best to pose a question. If you supply a data example and code, test for yourself that it runs as posted. If I tack the code you just provided onto the code in your first post I get a different error.

[Richmond Baye]

Thank you Devra,
This is the error I got when i tried estimating the model.

* Example generated by -dataex-. For more info, type help dataex
clear
input double(cost indemnity diseaseuncertainty) str17 responseid long choice str7 alternative2
1.5               .75 0 "R_12c0WhpSacmBfND" 3 "choiceA"
1.5               .75 0 "R_12c0WhpSacmBfND" 2 "choiceB"
1.5               .75 0 "R_12c0WhpSacmBfND" 3 "choiceC"
  0               .75 1 "R_130tjeDa0SB0a7n" 2 "choiceA"
  0               .75 1 "R_130tjeDa0SB0a7n" 2 "choiceB"
  0               .75 1 "R_130tjeDa0SB0a7n" 4 "choiceC"
  3 .8999999761581421 0 "R_17ro9DXzTCsyb4d" 3 "choiceA"
  3 .8999999761581421 0 "R_17ro9DXzTCsyb4d" 4 "choiceB"
  3 .8999999761581421 0 "R_17ro9DXzTCsyb4d" 4 "choiceC"
  3 .8999999761581421 0 "R_1B19p1QowTgqSm5" 3 "choiceA"
end
label values diseaseuncertainty diseaseuncertainty
label def diseaseuncertainty 0 "low", modify
label def diseaseuncertainty 1 "high", modify
label values choice choice
label def choice 2 "1", modify
label def choice 3 "2", modify
label def choice 4 "3", modify

------------------ copy up to and including the previous line ------------------

Listed 10 out of 231 observations

. cmmprobit choice cost indemnity diseaseuncertainty
depvar choice must be a 0/1 variable indicating which alternatives are chosen

[Devra Golbe]

I can't help thinking you've not shown us exactly [FAQ 12.1] what you typed, because when I type the code you specified in the code block and then the cmmprobit command, the error I get is

cmmprobit choice cost indemnity diseaseuncertainty
data not cmset
    Use cmset caseidvar ....
r(459);

I highly recommend following FAQ 12 in full, especially the part about showing a reproducible example that demonstrates your problem. If you can do so using a publicly available dataset (for instance, the travel dataset used in the cmmprobit manual entry), all the better.

[Joerg Luedicke (StataCorp)]

As the error message suggests, the variable that indicates chosen alternatives has to be a binary variable that takes on the value 1 if an alternative is chosen, and 0 otherwise. This is not true for the variable choice in your dataset.