Post estimation with meoprobit

[Eddy Simms]

* Example generated by -dataex-. For more info, type help dataex
clear
input byte pid double trt byte(gp score)
11 3 1 95
12 3 2 95
13 3 3 85
14 3 4 95
15 3 5 75
16 4 1 70
17 4 2 90
18 4 3 70
19 4 4 81
20 4 5 15
21 5 1 85
22 5 2 80
23 5 3 99
24 5 4 85
25 5 5 11
26 6 1 31
27 6 2 70
28 6 3 27
29 6 4 71
30 6 5  7
31 7 1 21
32 7 2 89
33 7 3 21
34 7 4 62
35 7 5  4
end

I have a repeated measures design where 5 GPs have scored 5 treatments on a 100mm VAS scale and I thought that meoprobit would be an appropriate method of analysis:

. meoprobit score i.trt || gp:,nolog

Mixed-effects oprobit regression                Number of obs     =         25
Group variable: gp                              Number of groups  =          5

                                                Obs per group:
                                                              min =          5
                                                              avg =        5.0
                                                              max =          5

Integration method: mvaghermite                 Integration pts.  =          7

                                                Wald chi2(4)      =      18.20
Log likelihood = -57.179953                     Prob > chi2       =     0.0011
------------------------------------------------------------------------------
       score | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
         trt |
          3  |      0.000  (base)
          4  |     -2.001      0.739    -2.71   0.007       -3.450      -0.552
          5  |     -1.184      0.699    -1.70   0.090       -2.553       0.185
          6  |     -3.244      0.872    -3.72   0.000       -4.952      -1.535
          7  |     -3.527      0.895    -3.94   0.000       -5.282      -1.772
-------------+----------------------------------------------------------------
       /cut1 |     -6.226      1.541                        -9.246      -3.206
       /cut2 |     -5.271      1.345                        -7.906      -2.635
       /cut3 |     -4.641      1.203                        -6.999      -2.283
       /cut4 |     -4.199      1.129                        -6.413      -1.986
       /cut5 |     -3.480      1.035                        -5.509      -1.452
       /cut6 |     -3.188      1.006                        -5.160      -1.216
       /cut7 |     -2.909      0.978                        -4.826      -0.993
       /cut8 |     -2.630      0.948                        -4.488      -0.772
       /cut9 |     -1.932      0.890                        -3.676      -0.188
      /cut10 |     -1.710      0.872                        -3.419      -0.001
      /cut11 |     -1.442      0.851                        -3.111       0.227
      /cut12 |     -1.188      0.840                        -2.834       0.457
      /cut13 |     -0.971      0.831                        -2.600       0.657
      /cut14 |     -0.373      0.816                        -1.973       1.226
      /cut15 |     -0.141      0.817                        -1.741       1.460
      /cut16 |      0.195      0.810                        -1.392       1.782
      /cut17 |      1.291      0.859                        -0.392       2.974
-------------+----------------------------------------------------------------
gp           |
   var(_cons)|      1.866      1.539                         0.370       9.401
------------------------------------------------------------------------------
LR test vs. oprobit model: chibar2(01) = 11.95        Prob >= chibar2 = 0.0003

I can see that there is an overall effect of treatment:

. testparm i(3/7).trt

 ( 1)  [score]4.trt = 0
 ( 2)  [score]5.trt = 0
 ( 3)  [score]6.trt = 0
 ( 4)  [score]7.trt = 0

           chi2(  4) =   18.20
         Prob > chi2 =    0.0011

but I am uncertain about comparing the individual treatments. Is pwcompare appropriate as probabilities are involved:

. pwcompare trt, groups

Pairwise comparisons of marginal linear predictions

Margins: asbalanced

-------------------------------------------------
             |                         Unadjusted
             |     Margin   Std. err.      groups
-------------+-----------------------------------
score        |
         trt |
          3  |      0.000      0.000            D
          4  |     -2.001      0.739          BC
          5  |     -1.184      0.699           CD
          6  |     -3.244      0.872         AB  
          7  |     -3.527      0.895         A   
-------------------------------------------------
Note: Margins sharing a letter in the group label
      are not significantly different at the 5%
      level.

Thank you.

Eddy Stata IC 17.0