why standard errors, z value, p-value dotted in regoprob2 with autofit estimation

[Abebe Desta]

dear all statalisters,
I have estimated the random effect generalized ordered probit model with an autofit option. i see a warning saying 'Warning: variance matrix is nonsymmetric or highly singular'
Besides, the result displayed all the coefficeints but the standard errors, p-values and others were dotted. what could be the problem? and how I can overcome this problem?

I am kindly asking your cooperation
thanks

--

----------------------------------------------------------------------------
Testing the parallel lines assumption using the .05 level of significance...

Step  1:  Constraints for parallel lines imposed for _Iregion_3 (P Value = 0.9680)
Step  2:  Constraints for parallel lines imposed for hh_s11aq03 (P Value = 0.9624)
Step  3:  Constraints for parallel lines imposed for flood (P Value = 0.9025)
Step  4:  Constraints for parallel lines imposed for hh_s11aq01 (P Value = 0.8364)
Warning:  variance matrix is nonsymmetric or highly singular
Step  5:  Constraints for parallel lines imposed for percapitaland (P Value = 0.8142)
Step  6:  Constraints for parallel lines imposed for hh_s1q04_a (P Value = 0.7936)
Warning:  variance matrix is nonsymmetric or highly singular
Step  7:  Constraints for parallel lines imposed for illnesshhmem (P Value = 0.7884)
Warning:  variance matrix is nonsymmetric or highly singular
Warning:  variance matrix is nonsymmetric or highly singular
Step  8:  Constraints for parallel lines are not imposed for
          hh_s1q03 (P Value =       .)
          maried (P Value =       .)
          _Ieduchead_1 (P Value =       .)
          _Ieduchead_2 (P Value =       .)
          _Ieduchead_3 (P Value =       .)
          _Ieduchead_4 (P Value =       .)
          _Ieduchead_5 (P Value =       .)
          adeqfamsize (P Value =       .)
          depratio (P Value =       .)
          agesquare (P Value =       .)
          femmembers (P Value =       .)
          TLU (P Value =       .)
          deathhhmem (P Value =       .)
          drought (P Value =       .)
          cropdamage (P Value =       .)
          loss_livestock (P Value =       .)
          Praisefooditem (P Value =       .)
          Incprice_input (P Value =       .)
          losshouse_farm (P Value =       .)
          hh_s12q02 (P Value =       .)
          hh_s11aq02 (P Value =       .)
          hh_s14q01 (P Value =       .)
          hh_s4q31 (P Value =       .)
          _Iregion_4 (P Value =       .)
          _Iregion_7 (P Value =       .)
          _Iregion_8 (P Value =       .)

Wald test of parallel lines assumption for the final model:

 ( 1)  [mleq1]_Iregion_3 - [mleq2]_Iregion_3 = 0
 ( 2)  [mleq1]hh_s11aq03 - [mleq2]hh_s11aq03 = 0
 ( 3)  [mleq1]flood - [mleq2]flood = 0
 ( 4)  [mleq1]hh_s11aq01 - [mleq2]hh_s11aq01 = 0
 ( 5)  [mleq1]percapitaland - [mleq2]percapitaland = 0
 ( 6)  [mleq1]hh_s1q04_a - [mleq2]hh_s1q04_a = 0
 ( 7)  [mleq1]illnesshhmem - [mleq2]illnesshhmem = 0
 ( 8)  [mleq1]_Iregion_3 - [mleq3]_Iregion_3 = 0
 ( 9)  [mleq1]hh_s11aq03 - [mleq3]hh_s11aq03 = 0
 (10)  [mleq1]flood - [mleq3]flood = 0
 (11)  [mleq1]hh_s11aq01 - [mleq3]hh_s11aq01 = 0
 (12)  [mleq1]percapitaland - [mleq3]percapitaland = 0
 (13)  [mleq1]hh_s1q04_a - [mleq3]hh_s1q04_a = 0
 (14)  [mleq1]illnesshhmem - [mleq3]illnesshhmem = 0

           chi2( 14) =    2.06
         Prob > chi2 =    0.9999

An insignificant test statistic indicates that the final model
does not violate the parallel lines assumption

If you re-estimate this exact same model with regoprob2, without
the option  autofit you can save time by using the parameter

pl(_Iregion_3 hh_s11aq03 flood hh_s11aq01 percapitaland hh_s1q04_a illnesshhmem)

------------------------------------------------------------------------------

Random Effects Generalized Ordered Probit         Number of obs   =       9584
                                                  Wald chi2(0)    =          .
Log likelihood = -9736.5841                       Prob > chi2     =          .

 ( 1)  [mleq1]_Iregion_3 - [mleq2]_Iregion_3 = 0
 ( 2)  [mleq1]hh_s11aq03 - [mleq2]hh_s11aq03 = 0
 ( 3)  [mleq1]flood - [mleq2]flood = 0
 ( 4)  [mleq1]hh_s11aq01 - [mleq2]hh_s11aq01 = 0
 ( 5)  [mleq1]percapitaland - [mleq2]percapitaland = 0
 ( 6)  [mleq1]hh_s1q04_a - [mleq2]hh_s1q04_a = 0
 ( 7)  [mleq1]illnesshhmem - [mleq2]illnesshhmem = 0
 ( 8)  [mleq2]_Iregion_3 - [mleq3]_Iregion_3 = 0
 ( 9)  [mleq2]hh_s11aq03 - [mleq3]hh_s11aq03 = 0
 (10)  [mleq2]flood - [mleq3]flood = 0
 (11)  [mleq2]hh_s11aq01 - [mleq3]hh_s11aq01 = 0
 (12)  [mleq2]percapitaland - [mleq3]percapitaland = 0
 (13)  [mleq2]hh_s1q04_a - [mleq3]hh_s1q04_a = 0
 (14)  [mleq2]illnesshhmem - [mleq3]illnesshhmem = 0
--------------------------------------------------------------------------------
   povdynamics |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
mleq1          |
      hh_s1q03 |   .0746513          .        .       .            .           .
    hh_s1q04_a |  -.0135047  (constrained)
        maried |   .0910478          .        .       .            .           .
  _Ieduchead_1 |   .2538072          .        .       .            .           .
  _Ieduchead_2 |   .5345384          .        .       .            .           .
  _Ieduchead_3 |   1.462105          .        .       .            .           .
  _Ieduchead_4 |   .6478492          .        .       .            .           .
  _Ieduchead_5 |  -.5137844          .        .       .            .           .
   adeqfamsize |  -.1672651          .        .       .            .           .
      depratio |  -.0653924          .        .       .            .           .
     agesquare |    .000135          .        .       .            .           .
    femmembers |  -.0158422          .        .       .            .           .
           TLU |   .0159989          .        .       .            .           .
 percapitaland |   .0097449  (constrained)
  illnesshhmem |   .0164412  (constrained)
    deathhhmem |   .3700783          .        .       .            .           .
       drought |   .0672083          .        .       .            .           .
    cropdamage |   .1088857          .        .       .            .           .
         flood |  -.1768249  (constrained)
loss_livestock |  -.0016499          .        .       .            .           .
Praisefooditem |   .0320447          .        .       .            .           .
Incprice_input |   .0468258          .        .       .            .           .
losshouse_farm |  -.2615648          .        .       .            .           .
     hh_s12q02 |   .0000327          .        .       .            .           .
    hh_s11aq01 |   .2662988  (constrained)
    hh_s11aq02 |   .0024192          .        .       .            .           .
    hh_s11aq03 |   .1683822  (constrained)
     hh_s14q01 |  -.0274438          .        .       .            .           .
      hh_s4q31 |  -.0499082          .        .       .            .           .
    _Iregion_3 |  -1.102472  (constrained)
    _Iregion_4 |   .1699791          .        .       .            .           .
    _Iregion_7 |  -.6619873          .        .       .            .           .
    _Iregion_8 |  -.0587307          .        .       .            .           .
         _cons |   3.684069          .        .       .            .           .
---------------+----------------------------------------------------------------
mleq2          |
      hh_s1q03 |   .1131231          .        .       .            .           .
    hh_s1q04_a |  -.0135047  (constrained)
        maried |   .0790267          .        .       .            .           .
  _Ieduchead_1 |   .1867171          .        .       .            .           .
  _Ieduchead_2 |    .495213          .        .       .            .           .
  _Ieduchead_3 |   .6908221          .        .       .            .           .
  _Ieduchead_4 |   .6817216          .        .       .            .           .
  _Ieduchead_5 |   .2240866          .        .       .            .           .
   adeqfamsize |   -.180721          .        .       .            .           .
      depratio |  -.1705295          .        .       .            .           .
     agesquare |   .0001443          .        .       .            .           .
    femmembers |    .015161          .        .       .            .           .
           TLU |   .0145117          .        .       .            .           .
 percapitaland |   .0097449  (constrained)
  illnesshhmem |   .0164412  (constrained)
    deathhhmem |   .2122116          .        .       .            .           .
       drought |   .0612989          .        .       .            .           .
    cropdamage |   .0220791          .        .       .            .           .
         flood |  -.1768249  (constrained)
loss_livestock |   .0817488          .        .       .            .           .
Praisefooditem |  -.1540323          .        .       .            .           .
Incprice_input |    .092898          .        .       .            .           .
losshouse_farm |   .1217267          .        .       .            .           .
     hh_s12q02 |   .0000121          .        .       .            .           .
    hh_s11aq01 |   .2662988  (constrained)
    hh_s11aq02 |    .083798          .        .       .            .           .
    hh_s11aq03 |   .1683822  (constrained)
     hh_s14q01 |  -.0650839          .        .       .            .           .
      hh_s4q31 |   -.052392          .        .       .            .           .
    _Iregion_3 |  -1.102472  (constrained)
    _Iregion_4 |   .0791283          .        .       .            .           .
    _Iregion_7 |  -.5929112          .        .       .            .           .
    _Iregion_8 |  -.0514474          .        .       .            .           .
         _cons |   2.545842          .        .       .            .           .
---------------+----------------------------------------------------------------
mleq3          |
      hh_s1q03 |  -.1012581          .        .       .            .           .
    hh_s1q04_a |  -.0135047  (constrained)
        maried |   .1934749          .        .       .            .           .
  _Ieduchead_1 |   .1549342          .        .       .            .           .
  _Ieduchead_2 |   .3286928          .        .       .            .           .
  _Ieduchead_3 |   .6412192          .        .       .            .           .
  _Ieduchead_4 |   .8364169          .        .       .            .           .
  _Ieduchead_5 |   .5334819          .        .       .            .           .
   adeqfamsize |  -.1692178          .        .       .            .           .
      depratio |  -.1916654          .        .       .            .           .
     agesquare |   .0001135          .        .       .            .           .
    femmembers |  -.0286391          .        .       .            .           .
           TLU |   .0058197          .        .       .            .           .
 percapitaland |   .0097449  (constrained)
  illnesshhmem |   .0164412  (constrained)
    deathhhmem |   .1187462          .        .       .            .           .
       drought |  -.0163121          .        .       .            .           .
    cropdamage |   .0294645          .        .       .            .           .
         flood |  -.1768249  (constrained)
loss_livestock |  -.0098203          .        .       .            .           .
Praisefooditem |  -.0434374          .        .       .            .           .
Incprice_input |   .2220039          .        .       .            .           .
losshouse_farm |   .4200302          .        .       .            .           .
     hh_s12q02 |   .0000124          .        .       .            .           .
    hh_s11aq01 |   .2662988  (constrained)
    hh_s11aq02 |   .0520616          .        .       .            .           .
    hh_s11aq03 |   .1683822  (constrained)
     hh_s14q01 |   .0126872          .        .       .            .           .
      hh_s4q31 |   .0359536          .        .       .            .           .
    _Iregion_3 |  -1.102472  (constrained)
    _Iregion_4 |   .1632844          .        .       .            .           .
    _Iregion_7 |  -.5381809          .        .       .            .           .
    _Iregion_8 |   .0402397          .        .       .            .           .
         _cons |   1.028728          .        .       .            .           .
---------------+----------------------------------------------------------------
rho            |
         _cons |   .5836591          .        .       .            .           .
--------------------------------------------------------------------------------

[Abebe Desta]

hello every one,
any one who can help please!!

regards
Abebe

[Phil Bromiley]

This is a user written program with relatively narrow usage. I'm not sure what the parallel lines operator does - if it is just imposing a few constraints, then that would seem unlikely to mess things up. The program's capability may be subsumed in more recent Stata versions.

My guess is that that your variables result in a fit so good that there is no variance left - "Warning: variance matrix is nonsymmetric or highly singular". In such a case, there is no way to estimate standard errors etc. This extreme level of fit with such a large sample probably means that something untoward is going on - it is highly unlikely that you've just found a set of variables that fit that well. I'd look at the correlation matrix to see if anything correlates unreasonably highly with the dv. I'd check the number of observations for each level of the dv. I'd think about running partial models to see where it breaks.

I suggest you look at the gologit2 as an alternative estimator. Williams' website also has some extremely helpful documentation to assist in understanding gologit (and many other statistical issues).

[Abebe Desta]

thanks phil for your kind respose. can I apply a gologit2 estimator while the sigma2_u is strongly significant in an xtprobit or logit model for a longitudnal data? the constrained regression has given me the following result. is it possible to apply gologit2 given such result? sorry, I am using it for the first time and do not have detail knowledge of such models.
the gologit2 estimation has also given me very low psuedo-squared (0.0716). does this model make sense?

Random-effects ordered probit regression        Number of obs      =      9584
Group variable: hh_id                           Number of groups   =      3203

Random effects u_i ~ Gaussian                   Obs per group: min =         2
                                                               avg =       3.0
                                                               max =         3

Integration method: mvaghermite                 Integration points =        20

                                                Wald chi2(33)      =    713.30
Log likelihood  = -9773.5513                    Prob > chi2        =    0.0000

----------------------------------------------------------------------------------
     povdynamics |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------+----------------------------------------------------------------
        hh_s1q03 |  -.0020538   .0680832    -0.03   0.976    -.1354945    .1313869
      hh_s1q04_a |  -.0141253   .0083524    -1.69   0.091    -.0304957    .0022452
       agesquare |   .0001319   .0000801     1.65   0.099     -.000025    .0002889
        hh_saq09 |   -.012422   .0175416    -0.71   0.479    -.0468028    .0219589
     adeqfamsize |  -.1598475   .0213027    -7.50   0.000       -.2016    -.118095
        depratio |  -.1707046   .0391832    -4.36   0.000    -.2475023    -.093907
      femmembers |  -.0070409   .0193868    -0.36   0.716    -.0450383    .0309565
          maried |   .1360285   .0517016     2.63   0.009     .0346952    .2373619
    _Ieduchead_1 |   .1759754   .0529622     3.32   0.001     .0721713    .2797794
    _Ieduchead_2 |    .400981   .0666162     6.02   0.000     .2704157    .5315463
    _Ieduchead_3 |   .6965844   .1524345     4.57   0.000     .3978183    .9953506
    _Ieduchead_4 |   .7942848   .1374944     5.78   0.000     .5248007    1.063769
    _Ieduchead_5 |   .2793383   .3483249     0.80   0.423    -.4033659    .9620426
             TLU |    .006967   .0020698     3.37   0.001     .0029103    .0110236
   percapitaland |    .009539   .0046234     2.06   0.039     .0004773    .0186008
    illnesshhmem |   .0150489   .0422301     0.36   0.722    -.0677205    .0978183
      deathhhmem |   .1816955   .0995844     1.82   0.068    -.0134864    .3768774
         drought |   .0217561   .0419394     0.52   0.604    -.0604437    .1039558
      cropdamage |   .0371355   .0626677     0.59   0.553    -.0856909    .1599619
           flood |  -.1757141   .0922048    -1.91   0.057    -.3564322    .0050039
  loss_livestock |   .0283767   .0456406     0.62   0.534    -.0610772    .1178307
  Praisefooditem |  -.0732162   .0431444    -1.70   0.090    -.1577777    .0113452
  Incprice_input |   .1529851      .0461     3.32   0.001     .0626307    .2433395
       hh_s12q02 |    .000013   4.72e-06     2.76   0.006     3.79e-06    .0000223
      hh_s11aq01 |   .2622448   .0701492     3.74   0.000     .1247549    .3997348
      hh_s11aq02 |   .0542145   .0735364     0.74   0.461    -.0899143    .1983432
      hh_s11aq03 |   .1644303    .085826     1.92   0.055    -.0037856    .3326463
       hh_s14q01 |  -.0213226   .0361397    -0.59   0.555    -.0921551    .0495098
        hh_s4q31 |  -.0123214   .0594009    -0.21   0.836    -.1287452    .1041023
      _Iregion_3 |  -1.090013   .0959089   -11.37   0.000    -1.277991   -.9020345
      _Iregion_4 |   .1345331   .0990019     1.36   0.174     -.059507    .3285733
      _Iregion_7 |  -.5764657   .0937278    -6.15   0.000    -.7601688   -.3927626
      _Iregion_8 |   .0073159   .0959985     0.08   0.939    -.1808377    .1954695
-----------------+----------------------------------------------------------------
           /cut1 |   -3.85213   .2272738   -16.95   0.000    -4.297578   -3.406681
           /cut2 |   -2.66671   .2253204   -11.84   0.000     -3.10833    -2.22509
           /cut3 |  -.9578657   .2231461    -4.29   0.000    -1.395224   -.5205074
-----------------+----------------------------------------------------------------
       /sigma2_u |   1.406428    .070176                      1.275397     1.55092
----------------------------------------------------------------------------------
LR test vs. oprobit regression:  chibar2(01) =  1908.12 Prob>=chibar2 = 0.0000

[Richard Williams]

You must rate. For some reason I can't get regoprob or regoprob2 to run at all anymore, even though they used to.

If you want a 2nd opinion, the free student edition of supermix will run mixed effects gologit models. It doesn't have anything like autofit, so you will have to constrain variables yourself. You can get it at

http://www.ssicentral.com/supermix/student.html