"margins" does not work for the "cmp" command (simultaneous equations)

[Woo Kim]

Hello,

I have a quick question regarding the -margins- after the -cmp- command. I am running a simultaneous equations (ordered probit) model but I cannot get the margins. There are two equations (one for PKG and the other for Trip) in my code. My model looks like this:

 cmp (PKG= Age Race_A Race_B Race_H Race_O Smartphone Income_25 Income_50 Income_75 Income_100 Child_1 Child_2) (Trip = PKG# Age Race_A Race_B Race_H Race_O Car_1 Car_2  Emp_Ft Weekend )  [pweight=allwt], indicators($cmp_oprobit $cmp_oprobit )

I got the model result and it looks okay.

Mixed-process regression
Number of obs = 2,160, Wald chi2(21) = 3582.08
Log pseudolikelihood = -3401.4193 , Prob > chi2 = 0.0000

	Coef.	Std. Err.	z	P>z	[95% Conf.	Interval]
PKG
Age	-0.00818	0.002182	-3.75	0	-0.01246	-0.00391
Race_A	0.003987	0.103061	0.04	0.969	-0.19801	0.205983
Race_B	-0.2102	0.084909	-2.48	0.013	-0.37662	-0.04378
Race_H	-0.08357	0.084729	-0.99	0.324	-0.24964	0.082497
Race_O	-0.51918	0.213183	-2.44	0.015	-0.93701	-0.10135
Smartphone	0.530897	0.10361	5.12	0	0.327825	0.733969
Income_25	0.163087	0.104023	1.57	0.117	-0.04079	0.366968
Income_50	0.393381	0.103228	3.81	0	0.191058	0.595705
Income_75	0.545501	0.112677	4.84	0	0.324659	0.766344
Income_100	0.757234	0.110492	6.85	0	0.540674	0.973795
Child_1	0.090436	0.078244	1.16	0.248	-0.06292	0.243791
Child_2	0.167188	0.079285	2.11	0.035	0.011792	0.322584
Trip
PKG#	0.44458	0.170289	2.61	0.009	0.11082	0.778341
Age	0.011561	0.004005	2.89	0.004	0.003712	0.01941
Race_A	-0.36778	0.245455	-1.5	0.134	-0.84886	0.113305
Race_B	0.280395	0.188232	1.49	0.136	-0.08853	0.649323
Race_H	0.146785	0.16831	0.87	0.383	-0.1831	0.476667
Race_O	-4.02057	0.677218	-5.94	0	-5.34789	-2.69325
Car_1	4.268134	0.341488	12.5	0	3.598829	4.937439
Car_2	4.430902	0.331977	13.35	0	3.780238	5.081566
Emp_Ft	-0.27855	0.121284	-2.3	0.022	-0.51626	-0.04084
Weekend	0.192648	0.134175	1.44	0.151	-0.07033	0.455626
/cut_1_1	-0.12556	0.166133	-0.76	0.45	-0.45117	0.200054
/cut_1_2	0.298658	0.163662	1.82	0.068	-0.02211	0.61943
/cut_1_3	1.25663	0.167456	7.5	0	0.928422	1.584838
/cut_1_4	1.628837	0.165322	9.85	0	1.304812	1.952862
/cut_2_1	6.616011	0.130187	50.82	0	6.360849	6.871174
/atanhrho_12	-0.64896	0.236971	-2.74	0.006	-1.11341	-0.1845
rho_12	-0.57097	0.159717			-0.80527	-0.18244

.
HOWEVER, -margins- does not work well for the second equation. This is the margins for the second question. All values are zeroes.
margins, dydx(*) predict(eq(#2) pr outcome(#1))

	dy/dx	Std. Err.	z	P>z	[95% Conf.	Interval]
Age	0	(omitted)
Race_A	0	(omitted)
Race_B	0	(omitted)
Race_H	0	(omitted)
Race_O	0	(omitted)
Smartphone	0	(omitted)
Income_25	0	(omitted)
Income_50	0	(omitted)
Income_75	0	(omitted)
Income_100	0	(omitted)
Child_1	0	(omitted)
Child_2	0	(omitted)

.
For the first equation, it works well for every outcome. For instance,
margins, dydx(*) predict(eq(#1) pr outcome(#1))

	dy/dx	Std. Err.	z	P>z	[95% Conf.	Interval]
Age	0.002593	0.00068	3.81	0	0.00126	0.003926
Race_A	-0.00126	0.032646	-0.04	0.969	-0.06525	0.062721
Race_B	0.066584	0.026922	2.47	0.013	0.013818	0.119351
Race_H	0.026472	0.026851	0.99	0.324	-0.02615	0.079098
Race_O	0.16446	0.067394	2.44	0.015	0.03237	0.296549
Smartphone	-0.16817	0.032171	-5.23	0	-0.23123	-0.10512
Income_25	-0.05166	0.03294	-1.57	0.117	-0.11622	0.012899
Income_50	-0.12461	0.032573	-3.83	0	-0.18845	-0.06077
Income_75	-0.1728	0.035361	-4.89	0	-0.2421	-0.10349
Income_100	-0.23987	0.034391	-6.97	0	-0.30727	-0.17246
Child_1	-0.02865	0.024763	-1.16	0.247	-0.07718	0.019888
Child_2	-0.05296	0.025014	-2.12	0.034	-0.10199	-0.00393
Car_1	0	(omitted)
Car_2	0	(omitted)
Emp_Ft	0	(omitted)
Weekend	0	(omitted)

.
Why am I getting all "zero"s in the marginal effects for the second equation?

I would really appreciate your help.

Sincerely,

[Phil Bromiley]

Welcome to Stata list. You will increase your chances of useful answer by following the FAQ on asking questions-provide Stata code in code delimiters, readable Stata output, and sample data using dataex. It is also a good idea to simplify your model and posting as much as you can while still reflecting the problem. Shorter posts are much more likely to get responses.

Does changing the order in which things appear in the CMP command change which equation has problems? If so, you might be able to get all the margins simply by changing the order of the equations. However, if this doesn't work, then you may need to contact the program's author. With user written programs, only the authors of the programs have often spent the time to really understand the programming details.

[Palee Hewage]

Read the cmp help file and try using " resultsform(structural | reduced) " in the code

here's the example in the cmp help file

cmp (price: priceO = quantity# pcompete income) (quantity: quantityO = price# praw), ind($cmp_oprobit $cmp_oprobit) nolr qui tech(dfp)
. cmp, resultsform(reduced)
. margins, dydx(praw) predict(outcome(3) eq(quantity) pr)