Small-Hsiao Test for IIA in multinomial logit ---------------------------------------------
^smhsiao^ ^,^ ^e^lim(yvalue) [ ^s^amp(sampvar) ^d^etail mlogit_options
^smhsiao^ conducts the Small-Hsiao test for Independence of Irrelevant Alternatives after a multinomial logit model.
The command displays the test statistic, degrees of freedom and P-value, and places values in the return array. ^return list^ for details.
The command uses the e(sample) defined in ^mlogit^ to define the observations over which the test is to be computed.
^elim(^#^)^ specifies the value of the dependent variable that should be excluded from the second-stage estimation. This is a required. Note that different test results can occur when different categories are excluded.
^samp(^varname^)^ specifies a variable that divides the data set into two half samples for estimation. This variable must take on two (and only two values). If it is not specified, the sample is divided randomly roughly in half.
^detail^ indicated that the various intermediate estimations should be displayed. By default they are not.
^mlogit_options^: any other valid options for the mlogit command, such as ^robust^, may be specified as well.
. ^smhsiao, elim(3) samp(sampvar)^
Estimates a multinomial logit model of pid, race and gender on vote, and then calculates the Small-Hsiao test of IIA using vote==3 as the eliminated category. The variable sampvar indicates the two half-samples.
Small, K. A. and C. Hsiao. 1985. "Multinomial Logit Specification Tests." _International Economic Review_ 26:619-27.
Zhang, Junsen and Saul D. Hoffman. 1993. "Discrete-Choice Logit Models." _Sociological Methods and Research_ 22(2):193-213.
Nick Winter Cornell University nwinter@@umich.edu
Also see --------
Manual: [R] mlogit, hausman On-line: help for @mlogit@, @hausman@