{smcl}
{* 25Nov2008}{...}
{cmd:help emlcclogit}
{hline}

{title:Title}

{p2colset 5 17 19 2}{...}
{p2col :{hi:emlcclogit}}EM latent class conditional logit model{p_end}
{p2colreset}{...}


{title:Syntax}

{p 8 15 2}
{cmd:emlcclogit}
{depvar}
[{indepvars}] {ifin} {cmd:,}
{cmdab:gr:oup(}{varname}{cmd:)}
[{opt cl:uster(varname)}
 {opt c:lasses(#)}
 {opt se:ed(#)}
 {opt maxit(#)}]
 {opt bsreps(#)}]

{title:Description}

{pstd}
{cmd:emlcclogit} fits latent class conditional logit models by using the Expectation Maximisation (EM) algorithm (Train, 2008). The data setup is the same as for {cmd:clogit}.


{title:Options for emlcclogit}

{phang}
{opth group(varname)} is required and specifies an identifier variable
for the choice occasions.

{phang}
{opth cluster(varname)} specifies an identifier variable for the choosers.  This option should be specified only when individuals make
several choices; i.e., the dataset is a panel.

{phang}
{opt classes(#)} specifies the number of separate classes for which shares and coefficients are to be estimated  The default is
{cmd:classes(2)}.

{phang}
{opt maxit(#)} specifies the maximum number of iterations of the EM algorithm to be performed before displaying the results if there is no convergence. The default is {cmd:maxit(50)}.

{phang}
{opt bsreps(#)} specifies the number of bootstrap samples to be taken for calculating standard errors.  The default is {cmd:bsreps(0)}, which calculates coefficients only.

{title:Examples}

{phang2}{cmd:emlcclogit choice speed cost, group(choiceid) cluster(cid) classes(5) bsreps(50)} {p_end}


{title:Reference}

{phang}Train, K. E. 2008. {it:EM Algorithms for Nonparametric Estimation of Mixing Distributions}.
Journal of Choice Modelling, 1(1), pp.40-69.


{title:Author}

{phang}This command was written by Paul Metcalfe (p.j.metcalfe@lse.ac.uk),
London School of Economics. Comments and suggestions are welcome. {p_end}


{title:Also see}

{psee}
Manual:  {bf:[R] clogit}

{psee}
Online:  {manhelp clogit R}{p_end}