Title
mtreatreg -- Multinomial treatment effects model
Syntax
mtreatreg depvar [indepvars] [if] [in] [weight], mtreatment(depvar_mt = indepvars_mt) simulationdraws(#) density(densityname) [options]
options Description ------------------------------------------------------------------------- Model mtreatment(string) string is specified as depvar_mt = indepvars_mt. It is required. simulationdraws(#) number of simulation draws per observation. It is required. density(densityname) distribution of depvar. It is required. prefix(string) prefix for the indicator variables created from the multinomial treatment variable. The default is a set of indicator variables starting with _T. constraint(constraint) apply specified linear constraints
SE/Robust vce(vcetype) vcetype may be oim, robust, opg, bootstrap or jackknife robust synonym for vce(robust) cluster(varname) adjust standard errors for intragroup correlation
Reporting verbose report mixed multinomial logit treatment and exogenous outcome regressions.
Quasi-random numbers facscale(#) specifies the standard deviation of each of the quasi-random variables. The default is 1. startpoint(#) specifies the starting point in the Halton sequence from which the quasi-random variates are generated. The default is 20.
Max options facfrom(string) specifies the starting value for the parameter associated with the latent factor. The default starting value is zero. maximize_options control the maximization process. These options are seldom used. -------------------------------------------------------------------------
densityname Description ------------------------------------------------------------------------- gamma Gamma logit Logistic negbin1 Negative Binomial-1 (constant dispersion) negbin2 Negative Binomial-2 (mean dispersion) normal Normal -------------------------------------------------------------------------
bootstrap, by, jackknife, rolling, statsby, and xi are allowed; see prefix. pweights, aweights, fweights, and iweights are allowed. See mtreatreg postestimation for features available after estimation.
Description
mtreatreg fits models with multinomial treatments and continuous, count and binary outcomes outcomes using maximum simulated likelihood. The model considers the effect of an endogenously chosen multinomial-valued treatment on an outcome variable, conditional on two sets of independent variables. The outcome variable can be continuous, binary or integer-valued while the treatment choice is assumed to follow a mixed multinomial logit distribution. The model is estimated using maximum simulated likelihood and the simulator uses Halton sequences.
Options
+-------+ ----+ Model +------------------------------------------------------------
mtreatment(depvar_mt = indepvars_mt{cmd:) specifies the variables for the treatment equation. It is required.
simulationdraws(#) specifies the number of simulation draws used per observation. It is required.
density(densityname) specifies the distribution of depvar. See densityname. It is required.
prefix(string) allows the user to choose a prefix other than _T for the indicator variables created from the multinomial treatment variable. The default is a set of indicator variables starting with _T. When mtreatreg is called, all previously created indicator variables starting with the prefix specified in the prefix(string) option or with _T are dropped.
noconstant, constraint(constraint), collinear; see estimation options.
+-----------+ ----+ SE/Robust +--------------------------------------------------------
vce(vcetype); see vce_option.
robust, cluster(varname); see estimation options.
+-----------+ ----+ Reporting +--------------------------------------------------------
verbose specifies that output from the mixed multinomial logit treatment and exogenous outcome regressions be reported. verbose has no effect if from is specified.
+----------------------+ ----+ Quasi-random numbers +---------------------------------------------
facscale(#) specifies the standard deviation of the quasi-random variables. The default is 1.
startpoint(#) specifies the starting point in the Halton sequence from which the quasi-random variates are generated. The default is 20.
+-------------+ ----+ Max options +------------------------------------------------------
facfrom(string)}specifies the starting value for the parameter associated with the latent factor. The default starting value is zero.
maximize_options: difficult, technique(algorithm_spec), iterate(#), [no]log, trace, gradient, showstep, hessian, shownrtolerance, tolerance(#), ltolerance(#), gtolerance(#), nrtolerance(#), nonrtolerance(#), from(init_specs); see maximize. These options are seldom used.
Remarks
difficult may sometimes be a useful option if convergence is slow.
robust produces standard errors which take simulation error into account. It is the preferred option unless cluster(varname) is specified.
The available outcome densities and associated conditional means are
Density mtreatreg option cond. mean ------------------------------------------------------------ Gamma density(gamma) exp(xb) Logistic density(logit) exp(xb)/(1+exp(xb)) Negative Binomial-1 density(negbin1) exp(xb) Negative Binomial-2 density(negbin2) exp(xb) Normal(Gaussian) density(normal) xb
Examples
. mtreatreg ycontinuous x1 x2, mtreat(d=x1 x2 z) sim(200) dens(gamma) . mtreatreg yinteger x1 x2, mtreat(d=x1 x2 z) sim(200) dens(negbin1)
References
Deb, P., and P. K. Trivedi (2006), Specification and Simulated Likelihood Estimation of a Non-normal Treatment-outcome Model with Selection: Application to Health Care Utilization, Econometrics Journal, 9, 307-331.
Deb, P., and P. K. Trivedi (2006), Maximum Simulated Likelihood Estimation of a Negative-binomial Regression Model with Multinomial Endogenous Treatment, The Stata Journal, 6, 246-255.
Author
Partha Deb, Hunter College and The Graduate Center, City University of New York, and NBER, USA. partha.deb@hunter.cuny.edu
Also see
Online: mtreatreg postestimation treatreg