Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Correlation estimates near boundary value in asmprobit


From   "Groeneveld, Rolf" <Rolf.Groeneveld@wur.nl>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: Correlation estimates near boundary value in asmprobit
Date   Sun, 18 May 2008 13:08:41 +0200

When using asmprobit to analyse the data of a stated preference survey I'm getting estimations of the correlation coefficient very close to 1 or -1. Different covariance matrices and different base and scale variables all give an estimate very close to its boundary. What could cause this result? Is it a problem for my estimates? What can I do about it?
 
Here is example of the results I get (the survey has three alternatives in each question):
 
. asmprobit choice stock benth layov costs, case(obs) alternatives(_altnum) base(1) scale(2)
 
Alternative-specific multinomial probit        Number of obs      =      22464
Case variable: obs                             Number of cases    =       7488
 
Alternative variable: _altnum                  Alts per case: min =          3
                                                              avg =        3.0
                                                              max =          3
Integration sequence:      Hammersley
Integration points:               150             Wald chi2(4)    =     423.47
Log simulated-likelihood = -6053.3294             Prob > chi2     =     0.0000
 
------------------------------------------------------------------------------
      choice |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
_altnum      |
       stock |   .1626189   .0162338    10.02   0.000     .1308011    .1944366
       benth |   .2973948   .0547786     5.43   0.000     .1900307     .404759
       layov |   .2402472   .0268319     8.95   0.000     .1876577    .2928367
       costs |  -.0138585   .0018745    -7.39   0.000    -.0175325   -.0101846
-------------+----------------------------------------------------------------
_alt_1       |  (base alternative)
-------------+----------------------------------------------------------------
_alt_2       |
       _cons |   .9968469   .1406466     7.09   0.000     .7211846    1.272509
-------------+----------------------------------------------------------------
_alt_3       |
       _cons |   1.060629   .1027839    10.32   0.000     .8591764    1.262082
-------------+----------------------------------------------------------------
     /lnl2_2 |  -2.930848   3.865245    -0.76   0.448    -10.50659    4.644894
-------------+----------------------------------------------------------------
       /l2_1 |   .9807269   .0573761    17.09   0.000     .8682719    1.093182
------------------------------------------------------------------------------
(_altnum=1 is the alternative normalizing location)
(_altnum=2 is the alternative normalizing scale)
 
. estat corr
 
  +-------------------------------+
  |              | _alt_2  _alt_3 |
  |--------------+----------------|
  |       _alt_2 |  1.000         |
  |       _alt_3 |  0.999   1.000 |
  +-------------------------------+
Note: correlations are for alternatives differenced with _alt_1 
 
Best regards,

Rolf A. Groeneveld
Wageningen University
Environmental Economics and Natural Resources Group
Hollandseweg 1
NL-6706 KN  Wageningen
+31 317 482009

<<winmail.dat>>




© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index