st: Trouble with mi impute truncreg

[Paul von Hippel <paulvonhippel.utaustin@gmail.com>]

In working with a simulated incomplete dataset, I find that I can
regress an incomplete x on y using the truncreg command, and I can
impute x conditionally on y using mi impute reg x y, add(5), but I
cannot impute the incomplete x conditionally on y using mi impute
truncreg, add(5). So truncreg works, and mi impute works, but they
don't work in combination. Many thanks for any ideas about why this
might be happening.

Details: I'm using the same dataset as in my previous message. (See
partial listing at the bottom of this message.) The truncreg command
converges just fine, but the mi impute truncreg command gives an error
that sounds like a failure to converge. Puzzling.

. truncreg x y, ll(0)
(note: 0 obs. truncated)

Fitting full model:

Iteration 0:   log likelihood = -13438.727
Iteration 1:   log likelihood = -12625.767
Iteration 2:   log likelihood = -12384.886
Iteration 3:   log likelihood = -12369.321
Iteration 4:   log likelihood = -12369.135
Iteration 5:   log likelihood = -12369.135

Truncated regression
Limit:   lower =          0                             Number of obs =   5950
        upper =       +inf                             Wald chi2(1)  =1943.64
Log likelihood = -12369.135                             Prob > chi2   = 0.0000

------------------------------------------------------------------------------
          x |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          y |   .8662065   .0196478    44.09   0.000     .8276975    .9047155
      _cons |  -5.099709    .322289   -15.82   0.000    -5.731384   -4.468034
-------------+----------------------------------------------------------------
     /sigma |   4.300881   .0863572    49.80   0.000     4.131624    4.470138
------------------------------------------------------------------------------

. mi impute truncreg x y, ll(0) add(5)
cannot compute an improvement -- discontinuous region encountered

. mi impute reg x y, add(5)

Univariate imputation                       Imputations =        5
Linear regression                                 added =        5
Imputed: m=1 through m=5                        updated =        0

------------------------------------------------------------------
                   |               Observations per m
                   |----------------------------------------------
          Variable |   Complete   Incomplete   Imputed |     Total
-------------------+-----------------------------------+----------
                 x |       5950         6050      6050 |     12000
------------------------------------------------------------------
(complete + incomplete = total; imputed is the minimum across m
 of the number of filled-in observations.)

. list x y

       +------------------------+
       |         x            y |
       |------------------------|
    1. | 2.2345289   -22.217398 |
    2. |         .   -21.833361 |
    3. | 4.6491754   -17.114244 |
    4. | 5.4836476   -16.483483 |
    5. |         .   -15.600014 |
  <snip>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/