[Limdep Nlogit List] Questions about LCA for ordered probit model

Thu Aug 29 10:34:36 AEST 2019

Hi! Can anyone help me with the following problem I met when doing LCA for
an ordered probit model?

"depen" is an ordered variable with the value range from 0 to 4.

I ran same codes for many times. Most of time, I got outputs like shown in
(1), but twice I got outputs like (2). Both the codes and data are exactly
the same!! Why did this happen? Should a start value be given to compute
function(Error 802)? How can I get results like (2) steadily? Many thanks!

(1) |-> ordered ; Lhs=depen

    ; Rhs=one, AGE,RACEWHIT, STUDENTS, MALE

    ; LCM=TRAVELWA, TRAVELCO, ENVENERG, LIKEDRIV,

    LIKERIDE, LIKETRAN, LIKEBIKE, LIKEWALK, WORRYOUT, LIKESHUT

    ; Pts=2$

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

Deleted     14 observations with missing data. N is now    162

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

Warning   43: ORDERED PROBIT - Current estimated thresholds  not ordered.

Error    802: Cannot compute function at start values. Exit status=4.

Function value was   .1951595121D+03 at entry -----------

                     .1581983858D+03 at exit  -----------

 (2)

|-> ordered ; Lhs=depen

    ; Rhs=one, AGE,RACEWHIT, STUDENTS, MALE

    ; LCM=TRAVELWA, TRAVELCO, ENVENERG, LIKEDRIV,

    LIKERIDE, LIKETRAN, LIKEBIKE, LIKEWALK, WORRYOUT, LIKESHUT

    ; Pts=2$

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

Deleted     14 observations with missing data. N is now    162

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

Warning   43: ORDERED PROBIT - Current estimated thresholds  not ordered.

Line search at iteration 33 does not improve the function

Exiting optimization

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

Latent Class / Panel OrdProbs Model

Dependent variable                DEPEN

Log likelihood function      -158.19839

Estimation based on N =    176, K =  27

Inf.Cr.AIC  =    370.4 AIC/N =    2.105

Sample is  1 pds and    176 individuals

Latent class model with 2 latent classes

Missing data: Skipped       14 individuals.

Ordered probability model

Ordered probit (normal) model

LHS variable = values 0,1,..., 4

--------+--------------------------------------------------------------------

        |                  Standard            Prob.      95% Confidence

   DEPEN|  Coefficient       Error       z    |z|>Z*         Interval

--------+--------------------------------------------------------------------

        |Model parameters for latent class
1.................................

Constant|    5.29996      .3857D+07      .00 1.0000 ***********  ***********

     AGE|     .08170         .12721      .64  .5208     -.16764    .33103

RACEWHIT|    -.23151         .86430     -.27  .7888    -1.92550   1.46248

STUDENTS|   -4.86029      .3857D+07      .00 1.0000 ***********  ***********

    MALE|   -2.08153        3.67508     -.57  .5711    -9.28456   5.12150

  Mu(01)|   -5.84096      .1474D+08      .00 1.0000 ***********  ***********

  Mu(02)|   -2.54470       14.21718     -.18  .8579   -30.40986  25.32045

  Mu(03)|    -.61230         .81213     -.75  .4509    -2.20406    .97945

        |Model parameters for latent class
2.................................

Constant|    4.85715***     1.19866     4.05  .0001     2.50781   7.20649

     AGE|    -.05742**       .02492    -2.30  .0212     -.10626   -.00858

RACEWHIT|    -.72129**       .31729    -2.27  .0230    -1.34317   -.09941
................................