{smcl}
{* *! version 1.0.0 3mar2014}{...}
{cmd: help switchoprobit}
{right:also see: {help switchoprobit postestimation}}
{hline}
{title:Ordered Probit Switching Regression}
{phang}
{bf:switchoprobit} {hline 2} Estimate effect of binary treatment on discrete, ordered outcome.
{title:Syntax}
{p 8 17 2}
{cmdab:switchoprobit}
y_ordered x_ordered
{ifin}
{weight}
{cmd:, treat}
(y_treat x_treat)
[{it: options}]
{synoptset 20 tabbed}{...}
{synopthdr}
{synoptline}
{syntab: Main}
{synopt:{opt c:luster(varname)}} cluster standard errors using {it:varname} {p_end}
{synopt:{opt r:obust}} compute robust variance covariance matrix / standard errors {p_end}
{synoptline}
{p 4 6 2}
{cmd: pweight, iweight, aweight}s are allowed; see {help weight}
{title:Description}
{pstd}
{cmd:switchoprobit} estimates a model in which {cmd:treat} is a binary indicator for a treatment ({it:y_treat}) for which selection is believed
correlated with the outcome of interest, {it: y_ordered}. The model assumes that the unobservables in treatment and outcome equations have a bivariate
normal distribution, and that outcomes for treated and untreated groups are distinct. (A test for the hypothesis that the treated and untreated groups
belong to a single outcome regime are reported as part of standard output.) Parameters of the model are estimated by maximum likelihood.
{title:Examples}
{phang} Let self assessed health {bf: SAH} be ordered on a 1-5 scale (excellent, very good, good, fair, poor), and {bf:medicaid} be an indicator of participation in Medicaid: {p_end}
{cmd:. use nhisdataex, clear}
{phang}{cmd:. switch_oprobit sah female married, treat(medicaid female married)}
{phang}{cmd:. switch_oprobit sah female married [pweight=weight], treat(medicaid female married) vce(robust)}
{smcl}
{title:Author}
{pstd}
Christian A. Gregory, Economic Research Service, USDA, cgregory@ers.usda.gov