mlogtest invalid syntax r(198);

[Hongjin Jo]

Dear all:

I am trying to diagnose the result of mlogit with mlogtest, iia command.

However, my stata says so:


mlogtest, iia

Hausman tests of IIA assumption (N=55373)

Ho: Odds(Outcome-J vs Outcome-K) are independent of other alternatives
invalid syntax
r(198);

I tried to google and find some relevant tip for this problem in statalist, but I could not figure out. Please help. I appreciate it.

If I try mlogtest, all, the result is below:


LR tests for independent variables (N=55373)

Ho: All coefficients associated with given variable(s) are 0

| chi2 df P>chi2
-----------------+-------------------------
V248_A | 121.808 3 0.000
V240 | 49.796 3 0.000
V239 | 169.668 3 0.000
V140 | 134.759 3 0.000
V141 | 7607.375 3 0.000
pts | 524.901 3 0.000
polity | 591.331 3 0.000

Wald tests for independent variables (N=55373)

Ho: All coefficients associated with given variable(s) are 0

| chi2 df P>chi2
-----------------+-------------------------
V248_A | 121.516 3 0.000
V240 | 49.730 3 0.000
V239 | 169.080 3 0.000
V140 | 136.656 3 0.000
V141 | 6295.052 3 0.000
pts | 517.168 3 0.000
polity | 560.894 3 0.000

Hausman tests of IIA assumption (N=55373)

Ho: Odds(Outcome-J vs Outcome-K) are independent of other alternatives
invalid syntax
r(198);

Thus, I think it stucks again at Hausman tests of IIA assumption.

Just in case, my analysis result is below:


mlogit V142_Z V248_A V240 V239 V140 V141 pts polity if good==1

Iteration 0: log likelihood = -69102.842
Iteration 1: log likelihood = -64560.092
Iteration 2: log likelihood = -64288.402
Iteration 3: log likelihood = -64287.149
Iteration 4: log likelihood = -64287.149

Multinomial logistic regression Number of obs = 55373
LR chi2(21) = 9631.39
Prob > chi2 = 0.0000
Log likelihood = -64287.149 Pseudo R2 = 0.0697

-------------------------------------------------------------------------------
V142_Z | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------+----------------------------------------------------------------
Very_Negative |
V248_A | -.0783596 .0236045 -3.32 0.001 -.1246236 -.0320956
V240 | -.0310157 .0312909 -0.99 0.322 -.0923448 .0303134
V239 | -.0575845 .0077785 -7.40 0.000 -.07283 -.0423389
V140 | .0254309 .0070425 3.61 0.000 .0116278 .0392341
V141 | -.3897272 .0066382 -58.71 0.000 -.4027378 -.3767166
pts | -.1652263 .0083955 -19.68 0.000 -.1816813 -.1487714
polity | .0226378 .0035672 6.35 0.000 .0156463 .0296293
_cons | 1.311785 .0917163 14.30 0.000 1.132025 1.491546
--------------+----------------------------------------------------------------
Negative |
V248_A | -.0030316 .015578 -0.19 0.846 -.0335639 .0275006
V240 | -.0771946 .0205735 -3.75 0.000 -.1175179 -.0368712
V239 | -.041215 .0051924 -7.94 0.000 -.051392 -.031038
V140 | .0237151 .0051551 4.60 0.000 .0136113 .033819
V141 | -.2031637 .0044679 -45.47 0.000 -.2119206 -.1944069
pts | -.0762023 .0052138 -14.62 0.000 -.0864212 -.0659833
polity | .0492637 .0024294 20.28 0.000 .0445022 .0540251
_cons | .5609802 .0631896 8.88 0.000 .4371308 .6848296
--------------+----------------------------------------------------------------
Positive | (base outcome)
--------------+----------------------------------------------------------------
Very_Positive |
V248_A | -.1993458 .0193022 -10.33 0.000 -.2371774 -.1615143
V240 | .1167159 .0255047 4.58 0.000 .0667276 .1667042
V239 | .0358473 .0063087 5.68 0.000 .0234824 .0482122
V140 | -.0567383 .0067511 -8.40 0.000 -.0699703 -.0435063
V141 | .1905826 .0064302 29.64 0.000 .1779798 .2031855
pts | -.0808074 .0065197 -12.39 0.000 -.0935858 -.068029
polity | -.0147761 .0025415 -5.81 0.000 -.0197573 -.0097948
_cons | -1.25555 .0787621 -15.94 0.000 -1.409921 -1.101179
-------------------------------------------------------------------------------


I am not sure how can I fix this. In fact, I tried ologit, but I could not find the brant test for the ologit as significant, that is why I am now using mlogit.

Thanks again for your help.

Best,
Hong Jin

[William Lisowski]

First, mlogtest is a user-supplied command not part of the distributed Stata installation. It is part of the spost package of commands, and that package exists in two versions, spost9 and spost13. From within Stata, search mlogtest brings up links to these packages. Clicking on the link for spost13, we see at the top of the output

DESCRIPTION/AUTHOR(S)
spost13_ado | SPost13 commands from Long and Freese (2014)
Regression Models for Categorical Outcomes using Stata, 3rd Edition.
Support www.indiana.edu/~jslsoc/spost.htm
Scott Long ([email protected]) & Jeremy Freese ([email protected])

Since your problem seems to be within mlogtest, or at a minimum, a very confusing error message from mlogtest, you would likely be best to contact the authors through the support URL.

Do take the time to be sure the version you have installed agrees with the version of Stata you are running.

[Richard Williams]

There are different versions of mlogtest around. Mine (most current version) is

. which mlogtest
c:\ado\plus\m\mlogtest.ado
*! version 3.5.0 2016-10-25 | long freese | problem in Stata 14.2 smhsiao

Notice it refers to a problem so if your version is older the problem may have been fixed.

Also, your output is very hard to read. Use code tags. See pt. 12 of the FAQ.