ordered probit in panel data

[pritika dua]

Hello
I am working on Stat 13.My dependent variable is a qualitative variable with values ranging from0-6. I plan to use ordered probit. I am not clear on the meaning of offset variable and how is should run this command?I ama novice in this area
thanks in anticipation

[Carlo Lazzaro]

Pritika:
welcome to the list.
I would start off with reading -probit- and -oprobit- entries in Stata .pdf manual.

[pritika dua]

Thank you sir for your prompt reply.
I am completely new to econometrics and am facing problem comprehending many terms.
I have gone through the section on panel data but under offset it is asking to mention varname.
I am not clear with this.

[Carlo Lazzaro]

Pritika:
the following Stata thread can be useful : https://www.stata.com/statalist/arch.../msg00082.html.
I would also advice you to familiarize yourself with -probit- and -oprobit- before moving to panel data analysis.
As an aside, please call me Carlo, as many on (and many more off) the list do.

[pritika dua]

Hello
I have tried marginal effects post xtoprobit using the mfx command in Stata13. I am getting the same coefficients as the original regression.Is something wong???Kindly guide

[Carlo Lazzaro]

Pritika:
why using the old-fashioned -mfx- when -margins- has widely superseded it?
Why not posting what you have typed and what Stata gave you back (as per FAQ) so that the interested listers can see what's going on with your analysis?

[pritika dua]

xtoprobit sum Notice CPCB MNC Union INNOVATIONAWARD PATENTFILED, offset(sum)

Fitting comparison model:

Iteration 0: log likelihood = -1723.0999
Iteration 1: log likelihood = -600.01555
Iteration 2: log likelihood = -535.4172
Iteration 3: log likelihood = -534.80069
Iteration 4: log likelihood = -534.79956
Iteration 5: log likelihood = -534.79956

Refining starting values:

Grid node 0: log likelihood = -529.66243

Fitting full model:

Iteration 0: log likelihood = -529.66243 (not concave)
Iteration 1: log likelihood = -519.59283
Iteration 2: log likelihood = -446.70456
Iteration 3: log likelihood = -441.5485
Iteration 4: log likelihood = -441.05758
Iteration 5: log likelihood = -441.05325
Iteration 6: log likelihood = -441.05332
Iteration 7: log likelihood = -441.05332

Random-effects ordered probit regression Number of obs = 897
Group variable: companynum Number of groups = 152

Random effects u_i ~ Gaussian Obs per group: min = 1
avg = 5.9
max = 6

Integration method: mvaghermite Integration points = 12

Wald chi2(6) = 73.20
Log likelihood = -441.05332 Prob > chi2 = 0.0000

( 1) [sum]_cons = 0
-----------------------------------------------------------------------------------
sum | Coef. Std. Err. z P>|z| [95% Conf. Interval]
------------------+----------------------------------------------------------------
Notice | .3042002 .1352312 2.25 0.024 .0391519 .5692486
CPCB | 1.175251 .2337227 5.03 0.000 .7171628 1.633339
MNC | 1.52322 .4189081 3.64 0.000 .7021752 2.344265
Union | .8157972 .2416081 3.38 0.001 .3422541 1.28934
INN | .1825001 .0833534 2.19 0.029 .0191303 .3458699
PAT | .0003482 .0010565 0.33 0.742 -.0017225 .0024189
_cons | 0 (omitted)
sum | 1 (offset)
------------------+----------------------------------------------------------------
/cut1 | .7191373 .2671277 2.69 0.007 .1955766 1.242698
/cut2 | 4.029229 .3247377 12.41 0.000 3.392755 4.665703
/cut3 | 6.549557 .3832683 17.09 0.000 5.798365 7.300749
/cut4 | 8.170945 .4410812 18.52 0.000 7.306442 9.035449
/cut5 | 9.953867 .5117506 19.45 0.000 8.950855 10.95688
/cut6 | 12.36265 .62221 19.87 0.000 11.14314 13.58216
------------------+----------------------------------------------------------------
/sigma2_u | 4.606139 1.01237 2.994015 7.08631
-----------------------------------------------------------------------------------
LR test vs. oprobit regression: chibar2(01) = 187.49 Prob>=chibar2 = 0.0000

. mfx

Marginal effects after meglm
y = Linear prediction (predict)
= 2.3933419
------------------------------------------------------------------------------
variable | dy/dx Std. Err. z P>|z| [ 95% C.I. ] X
---------+--------------------------------------------------------------------
Notice | .3042002 .13523 2.25 0.024 .039152 .569249 .050167
CPCB*| 1.175251 .23372 5.03 0.000 .717163 1.63334 .283166
MNC*| 1.52322 .41891 3.64 0.000 .702175 2.34426 .38796
Union*| .8157972 .24161 3.38 0.001 .342254 1.28934 .146042
INN | .1825001 .08335 2.19 0.029 .01913 .34587 .311037
PAT | .0003482 .00106 0.33 0.742 -.001723 .002419 12.0312
sum | (offset) 1.27425
sum | (offset1) 1.27425
------------------------------------------------------------------------------
(*) dy/dx is for discrete change of dummy variable from 0 to 1

[pritika dua]

Kindly guide

[pritika dua]

which margin command has to be used???

[Carlo Lazzaro]

Pritika:
you obtained exactly what you asked Stata for, as you can see from the following toy-example:

use http://www.stata-press.com/data/r14/margex
. regress y i.sex i.group age

      Source |       SS           df       MS      Number of obs   =     3,000
-------------+----------------------------------   F(4, 2995)      =    136.51
       Model |  214569.509         4  53642.3772   Prob > F        =    0.0000
    Residual |   1176863.5     2,995  392.942737   R-squared       =    0.1542
-------------+----------------------------------   Adj R-squared   =    0.1531
       Total |  1391433.01     2,999  463.965657   Root MSE        =    19.823

------------------------------------------------------------------------------
           y |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         sex |
     female  |   18.44069   .8819175    20.91   0.000     16.71146    20.16991
             |
       group |
          2  |   5.178636   .9584485     5.40   0.000     3.299352    7.057919
          3  |   13.45907   1.286127    10.46   0.000     10.93729    15.98085
             |
         age |  -.3298831   .0373191    -8.84   0.000    -.4030567   -.2567094
       _cons |   68.63586   1.962901    34.97   0.000     64.78709    72.48463
------------------------------------------------------------------------------

. margins, dydx(*)

Average marginal effects                        Number of obs     =      3,000
Model VCE    : OLS

Expression   : Linear prediction, predict()
dy/dx w.r.t. : 1.sex 2.group 3.group age

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         sex |
     female  |   18.44069   .8819175    20.91   0.000     16.71146    20.16991
             |
       group |
          2  |   5.178636   .9584485     5.40   0.000     3.299352    7.057919
          3  |   13.45907   1.286127    10.46   0.000     10.93729    15.98085
             |
         age |  -.3298831   .0373191    -8.84   0.000    -.4030567   -.2567094
------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

Then the main question is: what do you want to get from -margins-?

[pritika dua]

How to measure magnitude of estimated parameters?

[pritika dua]

I want to measure marginal effects at the average number of practices(sum) adopted and for one standard deviation change in continuous variable and a change from 0 to 1 for dummy variables

[Carlo Lazzaro]

Pritika:
see -help estat esize-.(for linear models only, though).
Otherwise, please provide more details about your query. Thanks.

[pritika dua]

Carlo I am measuring the impact of a number of organisational variable on the environmental management system(sum).EMS is a latent variable which represents the environmental management quality.After running the xtoprobit on my panel data I now want to measure the magnitude of impact of my Independent variables(both continous and dummy) on my latent dependent variable.

[Carlo Lazzaro]

Pritika:
for parameters interpretation, see -oprobit- entry in Stata .pdf manual.
If you want to calculate predicted probability of the dependent variable, see -help xtoprobit postestimation-, and related entry in Stata .pdf manual, Example 1.