Hierarchical ordered probit model

[sophie maene]

Hi,

I would love to estimate an hierarchical ordered probit model, but I have only found the command

meoprobit DV IV, vce (robust)

Nevertheless, I would love to exploit the longitudinal nature of my dataset. Is there an random effects variant? (such as the xtoprobit command for random effects ordered probit models?) I now I can do

meoprobit DV IV, re vce (robust)

but this is only for cross-sectional datasets.

[Andrew Musau]

Nevertheless, I would love to exploit the longitudinal nature of my dataset. Is there an random effects variant? (such as the xtoprobit command for random effects ordered probit models?)

I am not sure that I understand what you are getting at. xtprobit estimates a two-level random intercept model which can also be estimated using meprobit. If you want to estimate a two-level mixed effects ordered probit model, then go right ahead.

webuse union
xtset idcode year
xtprobit union age grade i.not_smsa south##c.year
meprobit union age grade i.not_smsa south##c.year|| idcode:

Res.:

. xtprobit union age grade i.not_smsa south##c.year

Fitting comparison model:

Iteration 0:   log likelihood =  -13864.23  
Iteration 1:   log likelihood = -13545.541  
Iteration 2:   log likelihood = -13544.385  
Iteration 3:   log likelihood = -13544.385  

Fitting full model:

rho =  0.0     log likelihood = -13544.385
rho =  0.1     log likelihood = -12237.655
rho =  0.2     log likelihood = -11590.282
rho =  0.3     log likelihood = -11211.185
rho =  0.4     log likelihood = -10981.319
rho =  0.5     log likelihood = -10852.793
rho =  0.6     log likelihood = -10808.759
rho =  0.7     log likelihood =  -10865.57

Iteration 0:   log likelihood = -10807.712  
Iteration 1:   log likelihood = -10599.332  
Iteration 2:   log likelihood = -10552.287  
Iteration 3:   log likelihood = -10552.225  
Iteration 4:   log likelihood = -10552.225  

Random-effects probit regression                Number of obs     =     26,200
Group variable: idcode                          Number of groups  =      4,434

Random effects u_i ~ Gaussian                   Obs per group:
                                                              min =          1
                                                              avg =        5.9
                                                              max =         12

Integration method: mvaghermite                 Integration pts.  =         12

                                                Wald chi2(6)      =     220.91
Log likelihood  = -10552.225                    Prob > chi2       =     0.0000

------------------------------------------------------------------------------
       union |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0082967   .0084599     0.98   0.327    -.0082843    .0248778
       grade |   .0482731   .0099469     4.85   0.000     .0287776    .0677686
  1.not_smsa |   -.139657   .0460548    -3.03   0.002    -.2299227   -.0493913
     1.south |  -1.584394    .358473    -4.42   0.000    -2.286989   -.8818002
        year |  -.0039854   .0088399    -0.45   0.652    -.0213113    .0133406
             |
south#c.year |
          1  |   .0134017   .0044622     3.00   0.003     .0046559    .0221475
             |
       _cons |  -1.668202   .4751819    -3.51   0.000    -2.599542   -.7368628
-------------+----------------------------------------------------------------
    /lnsig2u |   .6103616   .0458783                      .5204418    .7002814
-------------+----------------------------------------------------------------
     sigma_u |    1.35687   .0311255                      1.297217    1.419267
         rho |   .6480233   .0104643                      .6272511    .6682502
------------------------------------------------------------------------------
LR test of rho=0: chibar2(01) = 5984.32                Prob >= chibar2 = 0.000

. 
. meprobit union age grade i.not_smsa south##c.year|| idcode:

Fitting fixed-effects model:

Iteration 0:   log likelihood = -13560.682  
Iteration 1:   log likelihood = -13544.386  
Iteration 2:   log likelihood = -13544.385  

Refining starting values:

Grid node 0:   log likelihood = -10852.439

Fitting full model:

Iteration 0:   log likelihood = -10852.439  
Iteration 1:   log likelihood = -10582.047  
Iteration 2:   log likelihood = -10548.924  
Iteration 3:   log likelihood = -10548.759  
Iteration 4:   log likelihood = -10548.762  
Iteration 5:   log likelihood = -10548.762  

Mixed-effects probit regression                 Number of obs     =     26,200
Group variable:          idcode                 Number of groups  =      4,434

                                                Obs per group:
                                                              min =          1
                                                              avg =        5.9
                                                              max =         12

Integration method: mvaghermite                 Integration pts.  =          7

                                                Wald chi2(6)      =     226.73
Log likelihood = -10548.762                     Prob > chi2       =     0.0000
------------------------------------------------------------------------------
       union |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |    .008223   .0082976     0.99   0.322      -.00804    .0244861
       grade |   .0484164   .0098133     4.93   0.000     .0291827    .0676501
  1.not_smsa |  -.1392458   .0455343    -3.06   0.002    -.2284913   -.0500002
     1.south |   -1.58779   .3582428    -4.43   0.000    -2.289933   -.8856471
        year |  -.0039134   .0086835    -0.45   0.652    -.0209327    .0131058
             |
south#c.year |
          1  |   .0134334   .0044595     3.01   0.003     .0046928     .022174
             |
       _cons |   -1.68694   .4679579    -3.60   0.000    -2.604121   -.7697595
-------------+----------------------------------------------------------------
idcode       |
   var(_cons)|   1.890777   .0906564                      1.721187    2.077077
------------------------------------------------------------------------------
LR test vs. probit model: chibar2(01) = 5991.25       Prob >= chibar2 = 0.0000

[Richard Williams]

What do you have against using xtoprobit?

At least for simpler models, it is pretty easy to get meoprobit to do the same things xtoprobit does:

webuse tvsfpors, clear
xtset school
xtoprobit thk prethk cc##tv, nolog vce(robust)
meoprobit thk prethk cc##tv || school:, nolog vce(robust)

[Shadrack Mutembereza]

Richard Williams
What is the best specification for weight in xtoprobit?
svy: xtoprobit or
xtoprobit Y X1 X2 X2 [pweight=weight]

I don't understand the difference

[Richard Williams]

It definitely makes a difference if you have stratification or clustering. Also if you are analyzing subopulations. See

https://www.statalist.org/forums/for...-are-available

Even though I asked the original Q in that thread, I am still a little confused.

An added consideration in this specific case: I don't think svy: works with xtoprobit. But meoprobit does.

Anyway, I always just use svy:

[Shadrack Mutembereza]

svy: definitely doesn't work with XTOPROBIT.

I don't know if the error of "weight not constant within groups of defined by ID" can be addressed. I also would like to know the major cause of the error.