Urgent help - xtoprobit panel regression

[Abebe Desta]

I am using stata 13 and and regressing xtoprobit for a three round panel. the result gives me that the number of observations just equal to the number of groups. Is there posibility to have same observation and group numbers in a panel data? How can I correct it? . Sigma2_u is also almost zero implying there is no variation over time and I applied oprobit regression as a solution. the result still revealed that the pseudo R-square is very low (below ten percent) and the number of observations I obtained are equal to the number of groups (3143). where I made wrong? I couldnt identify it. Anyone to help me out please?

[Carlo Lazzaro]

Abebe:
please share what you typed and what Stata gave you back (via CODE delimiters, please) and/or post an excerpt/example of your dataset via -dataex-.
Eventually, crying for help/urgent matters, usually reduces your chances of getting an helpful reply: all listers are busy people trying to deal with (too) many engagements and the Stata forum is not intended as a customer service..

[Abebe Desta]

dear Carlo Lazzaro thanks for your valuable comment.
I run the the random effect ordered probit model (xtoprobit) . and the results are partially shown below. how can I correct this problem

xtset hh_id hh_saq13_c
panel variable: hh_id (strongly balanced)
time variable: hh_saq13_c, 2004 to 2008, but with gaps
delta: 1 unit

Random-effects ordered probit regression Number of obs = 3143
Group variable: hh_id Number of groups = 3143

Random effects u_i ~ Gaussian Obs per group: min = 1
avg = 1.0
max = 1

Integration method: mvaghermite Integration points = 12

Wald chi2(36) = 813.03
Log pseudolikelihood = -3875.6153 Prob > chi2 = 0.0000

(Std. Err. adjusted for 3143 clusters in hh_id)
----------------------------------------------------------------------------------
| Robust
povcategory | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
hh_s1q03 | -.0309128 .0667902 -0.46 0.643 -.1618192 .0999935
hh_s1q04_a | -.0074715 .0080556 -0.93 0.354 -.0232601 .0083171
_Imaried_1 | .221124 .0593326 3.73 0.000 .1048343 .3374138
_Ieduchead_1 | .2819786 .0575425 4.90 0.000 .1691974 .3947597
_Ieduchead_2 | .4326263 .0678121 6.38 0.000 .299717 .5655357
_Ieduchead_3 | .3255298 .1559916 2.09 0.037 .0197918 .6312677
_Ieduchead_4 | .6564651 .139211 4.72 0.000 .3836165 .9293137
_Ieduchead_5 | .4756074 .4164513 1.14 0.253 -.340622 1.291837
adeqfamsize | -.1608969 .0161005 -9.99 0.000 -.1924533 -.1293406
depratio | -.4610711 .09085 -5.08 0.000 -.6391339 -.2830083
agesquare | .000062 .0000793 0.78 0.434 -.0000934 .0002174
femmembers | -.0091839 .0198162 -0.46 0.643 -.048023 .0296551
TLU | .0200844 .0064792 3.10 0.002 .0073854 .0327833
percapitaland | .0029209 .0022646 1.29 0.197 -.0015177 .0073595
deathhhmem | .2151211 .109719 1.96 0.050 .0000757 .4301664
illnesshhmem | -.103488 .0581217 -1.78 0.075 -.2174043 .0104284
drought | -.1006613 .0569744 -1.77 0.077 -.2123291 .0110064
flood | -.4234966 .1072483 -3.95 0.000 -.6336994 -.2132938
....................
.................
-----------------+----------------------------------------------------------------
/cut1 | -2.204827 .206864 -10.66 0.000 -2.610273 -1.799381
/cut2 | -1.29916 .2059173 -6.31 0.000 -1.702751 -.8955698
/cut3 | -.401736 .2046614 -1.96 0.050 -.8028651 -.000607
-----------------+----------------------------------------------------------------
/sigma2_u | 9.93e-24 3.51e-23 9.84e-27 1.00e-20
----------------------------------------------------------------------------------

[William Lisowski]

If you only have observations in 2004, 2006, and 2008 you should add the delta option to your xtset command. See the output of help xtset for instructions.

[Abebe Desta]

thanks William. I will check that and write the feedback

[Carlo Lazzaro]

Abebe:
please use CODE delimiters in sharing Stata codes and output obtained with William's helpful fix.

[Abebe Desta]

dear William, I have applied the delta option and the extset command brought the following result. but the xtoprobit result still remained the same. neither the observations nor the significance of Sigma2_u are changed. even all the new coefficients are the same to the previous output. What else can I do to correct the model appreciating your comments guys?


xtset hh_id hh_saq13_c, delta(2)
panel variable: hh_id (strongly balanced)
time variable: hh_saq13_c, 2004 to 2008
delta: 2 units

[Abebe Desta]

Sorry Carlo, I failed to Send the stata results through the CODE delimiters. Can you give me a hint on how to use it?. I apologize for getting you back in to such preliminary tasks. it is because I am not that much experienced in stata.

[Carlo Lazzaro]

The #-shaped button in the Advanced editor toolbar gives you access to CODE delimiters. When enetered between the tags of CODE delimiters, your Stata codes/output/tables will look like this

[Abebe Desta]

Thanks Carlo. this is my xtoprobit result random effect model. waiting for your suggestions please!!
I thank you in advance

Random-effects ordered probit regression        Number of obs      =      3143
Group variable: hh_id                           Number of groups   =      3143

Random effects u_i ~ Gaussian                   Obs per group: min =         1
                                                               avg =       1.0
                                                               max =         1

Integration method: mvaghermite                 Integration points =        12

                                                Wald chi2(36)      =    813.03
Log pseudolikelihood  = -3875.6153              Prob > chi2        =    0.0000

                                   (Std. Err. adjusted for 3143 clusters in hh_id)

----------------------------------------------------------------------------------
                 |               Robust
     povcategory |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------+----------------------------------------------------------------
        hh_s1q03 |  -.0309128   .0667902    -0.46   0.643    -.1618192    .0999935
      hh_s1q04_a |  -.0074715   .0080556    -0.93   0.354    -.0232601    .0083171
      _Imaried_1 |    .221124   .0593326     3.73   0.000     .1048343    .3374138
    _Ieduchead_1 |   .2819786   .0575425     4.90   0.000     .1691974    .3947597
    _Ieduchead_2 |   .4326263   .0678121     6.38   0.000      .299717    .5655357
    _Ieduchead_3 |   .3255298   .1559916     2.09   0.037     .0197918    .6312677
    _Ieduchead_4 |   .6564651    .139211     4.72   0.000     .3836165    .9293137
    _Ieduchead_5 |   .4756074   .4164513     1.14   0.253     -.340622    1.291837
     adeqfamsize |  -.1608969   .0161005    -9.99   0.000    -.1924533   -.1293406
        depratio |  -.4610711     .09085    -5.08   0.000    -.6391339   -.2830083
       agesquare |    .000062   .0000793     0.78   0.434    -.0000934    .0002174
      femmembers |  -.0091839   .0198162    -0.46   0.643     -.048023    .0296551
             TLU |   .0200844   .0064792     3.10   0.002     .0073854    .0327833
   percapitaland |   .0029209   .0022646     1.29   0.197    -.0015177    .0073595
      deathhhmem |   .2151211    .109719     1.96   0.050     .0000757    .4301664
    illnesshhmem |   -.103488   .0581217    -1.78   0.075    -.2174043    .0104284
         drought |  -.1006613   .0569744    -1.77   0.077    -.2123291    .0110064
           flood |  -.4234966   .1072483    -3.95   0.000    -.6336994   -.2132938
      cropdamage |  -.1066065   .0908098    -1.17   0.240    -.2845905    .0713774
  Praisefooditem |   .0652509   .0497371     1.31   0.190     -.032232    .1627337
  Incprice_input |   .2254308   .0652071     3.46   0.001     .0976272    .3532344
  loss_livestock |  -.1438035   .0751706    -1.91   0.056    -.2911352    .0035282
       hh_s12q02 |   .0000444   9.35e-06     4.75   0.000      .000026    .0000627
    Ftransf_gift |  -.0000997   .0000996    -1.00   0.317     -.000295    .0000956
     inkind_gift |   .0000751    .000088     0.85   0.394    -.0000974    .0002476
      animalrent |   .0000607   .0000855     0.71   0.478     -.000107    .0002283
        non_agri |  -.0000911   .0000762    -1.20   0.232    -.0002404    .0000582
      hh_s11aq01 |   .1572737   .0877541     1.79   0.073    -.0147212    .3292686
      hh_s11aq02 |   .1180775   .0870711     1.36   0.175    -.0525786    .2887337
      hh_s11aq03 |   .1917233    .089442     2.14   0.032     .0164201    .3670264
       hh_s14q01 |  -.0571671   .0480951    -1.19   0.235    -.1514317    .0370976
        hh_s4q31 |  -.0646825   .0640756    -1.01   0.313    -.1902683    .0609034
      _Iregion_3 |  -.8869998   .0740548   -11.98   0.000    -1.032145    -.741855
      _Iregion_4 |    .083901   .0789604     1.06   0.288    -.0708584    .2386605
      _Iregion_7 |  -.4403021   .0761376    -5.78   0.000    -.5895289   -.2910752
      _Iregion_8 |    .030054   .0799873     0.38   0.707    -.1267182    .1868262
-----------------+----------------------------------------------------------------
           /cut1 |  -2.204827    .206864   -10.66   0.000    -2.610273   -1.799381
           /cut2 |   -1.29916   .2059173    -6.31   0.000    -1.702751   -.8955698
           /cut3 |   -.401736   .2046614    -1.96   0.050    -.8028651    -.000607
-----------------+----------------------------------------------------------------
       /sigma2_u |   9.93e-24   3.51e-23                      9.84e-27    1.00e-20
----------------------------------------------------------------------------------

[Abebe Desta]

thanks Carlo, this is my xtoprobit regression result after fixing the correction forwarded by William. Waiting your valuable comments please!!

[Carlo Lazzaro]

Abebe:
you does not seem to have panel data, but one-wave data only (ie, one observation per panel).
Try -oprobit-, instead.

[Abebe Desta]

that is also my problem. Why the model considers only one wave while It contains three-waves data? could there be any technical mistake i commited? I realy have a panel of 3201 households for three rounds. what possible reasons could be there for the model to treat only one wave?

thanks calrlo

[William Lisowski]

I believe you have one or more variables that are present in only one wave and missing in the other two waves. Stata will omit observations with missing values. That would explain why xtlogit shows one onservation for each of 3143 (not 3201) households.

[Abebe Desta]

hello every one
thanks all for your propt resposes. william's comment was right. I corrected a variable with missed information for two waves. But i encountered another problem. When I try to compute 'xtoprobit' it gives 'cannot compute an improvement- discontinuous region encountered'. what could be the reason for this and how could i correct this?

xi: xtoprobit povcategory hh_s1q03 hh_s1q04_a i.maried i.educhead adeqfamsize depratio agesquare  TLU percapitaland deathhhmem illnesshhmem drought  cropdamage Praisefooditem loss_livestock hh_s12q02 ctransfer_gift hh_s11aq01 hh_s11aq02 hh_s11aq03  hh_s14q01 hh_s4q33 i.region, vce(robust)
i.maried          _Imaried_0-1        (naturally coded; _Imaried_0 omitted)
i.educhead        _Ieduchead_0-5      (naturally coded; _Ieduchead_0 omitted)
i.region          _Iregion_1-8        (naturally coded; _Iregion_1 omitted)

Fitting comparison model:

Iteration 0:   log likelihood = -13096.341  
Iteration 1:   log likelihood = -11987.016  
Iteration 2:   log likelihood = -11981.487  
Iteration 3:   log likelihood = -11981.438  
Iteration 4:   log likelihood = -11981.437  

Refining starting values:

Grid node 0:   log likelihood = -10048.616

Fitting full model:

Iteration 0:   log pseudolikelihood = -10048.616  
Iteration 1:   log pseudolikelihood = -6217.8365  
adaptive quadrature failed to converge
cannot compute an improvement -- discontinuous region encountered
r(430);