-cmp- calculate margins of an IV ordered probit model

[Anna Mogilevskaja]

Hi,

I am estimating an IV ordered probit model using the cmp command. Afterwards, I am trying to calculate the marginal effects of my main regressors but I always get an error message.
The code looks like this:
cmp (PlannedDurables = inflexp income_exp~d i.expmacroqu~g $controls) (inflexp = devinfpoint income_exp~d i.expmacroqu~g $controls), ind($cmp_oprobit $cmp_cont) qui
margins, dydx(inflexp) predict(equation(PlannedDurables) pr) force

The error message is:
__marg_pvar_1 ambiguous abbreviation

Does anyone have an idea what causes the error message?
Thanks for the help.

[David Roodman]

I have managed to reproduce the problem and am working on it...

[David Roodman]

The margins command calls the predict command. predict in turn calls cmp_p.ado, which defines the behavior of predict for cmp. By default, when you run predict for an ordered-probit equation, cmp_p generates a set of variables, such as the probabilities for each possible outcome. This is confusing margins. To fix it, run the margins command for one possible outcome at a time. For example:

margins, dydx(inflexp) predict(equation(PlannedDurables) outcome(#1) pr) force
margins, dydx(inflexp) predict(equation(PlannedDurables) outcome(#2) pr) force
margins, dydx(inflexp) predict(equation(PlannedDurables) outcome(#3) pr) force

I'm optimistic this will work. For more on the outcome() option after ordered probit, see the very end of the cmp help file. In the outcome option, including a "#" lets you refer to the possible outcomes in order starting from 1. Leaving out the "#" lets you refer to their actual values. So if the actual possible values are -3, -2, -1 then outcome(#1) and outcome(-3) mean the same thing.

[Anna Mogilevskaja]

Thank you, David. It works now.

[Palee Hewage]

Hello, I am trying to calculate the marginal effects of the multivariate probit model (04 dependent binary variables) that I have developed using cmp
The code is mentioned below.

cmp (Y1 = X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X27) (Y2 = Y1# X15 X2 X4 X16 X17 X10 X18 X19 X14 X20) (Y3 = X15 X2 X3 X4 X21 X22 X8 X6 X7 X9 X23 X10 X12 X20) (Y4 =Y1# Y3# X2 X4 X8 X7 X24 X25 X10X11 X26 X28 X29 X30 X31 ),indicators($cmp_probit $cmp_probit $cmp_probit $cmp_probit) nolr redraws(200, anti) tech(dfp)


code margins, dydx(*) predict(pr) provide marginal effects with respect to Y1

But I could not calculate marginal effects with respect to Y2 with the code margins,dydx(*) predict(eq( Y2 ) pr) force

There, the Output for the dy/dx says " 0 (omitted)"

Is there anyone who could help me with the code, please?

Thanks for the help.