Hi everyone,
I am using cmp to run an ordered probit model with a right-censored dependent variable and complex data. I get different results in a Stata 11.2 (Mac) and Stata 12.1 (Windows), particularly in the arguments of the F test. Here a simplified version of the model with only one explanatory variable. Variable educ2 takes 3 values, it is censored for those still studying (stud==1), status and age1 are dummy variables.
Using Stata 11.2 in a Mac:
Using Stata 12.2 in a Windows:
In both computers, this is the version of the command :
If I remove the subpop() option or if I use $cmp_cont instead of $cmp_oprobit, the discrepancy disappears.
I appreciate very much your help.
Celia P.
I am using cmp to run an ordered probit model with a right-censored dependent variable and complex data. I get different results in a Stata 11.2 (Mac) and Stata 12.1 (Windows), particularly in the arguments of the F test. Here a simplified version of the model with only one explanatory variable. Variable educ2 takes 3 values, it is censored for those still studying (stud==1), status and age1 are dummy variables.
Using Stata 11.2 in a Mac:
Code:
. cmp (educ2 = status), ind("cond(stud==1, $cmp_right, $cmp_oprobit)") svy subpop(age1) qui Fitting individual models as starting point for full model fit. Note: For programming reasons, these initial estimates may deviate from your specification. For exact fits of each equation alone, run cmp separately on each. Fitting full model. Mixed-process regression Number of strata = 8 Number of obs = 46705 Number of PSUs = 4396 Population size = 14309427 Subpop. no. of obs = 10789 Subpop. size = 3602626.7 Design df = 4388 F( 0, 4388) = . Prob > F = . ------------------------------------------------------------------------------ | Linearized educ2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- educ2 | status | 1.09e+07 5593122 1.95 0.051 -60085.23 2.19e+07 -------------+---------------------------------------------------------------- /cut_1_1 | -1.28e+08 2469636 -51.94 0.000 -1.33e+08 -1.23e+08 /cut_1_2 | -5705095 2910510 -1.96 0.050 -1.14e+07 972.5065 /lnsig_1 | 18.70615 .0058611 3191.60 0.000 18.69466 18.71764 -------------+---------------------------------------------------------------- sig_1 | 1.33e+08 779749 1.32e+08 1.35e+08 ------------------------------------------------------------------------------
Using Stata 12.2 in a Windows:
Code:
. cmp (educ2 = status), ind("cond(stud==1, $cmp_right, $cmp_oprobit)") svy subpop(age1) qui Fitting individual models as starting point for full model fit. Note: For programming reasons, these initial estimates may deviate from your specification. For exact fits of each equation alone, run cmp separately on each. Fitting full model. Mixed-process regression Number of strata = 8 Number of obs = 46705 Number of PSUs = 4396 Population size = 14309427 Subpop. no. of obs = 10789 Subpop. size = 3602626.7 Design df = 4388 F( 1, 4388) = 3.80 Prob > F = 0.0513 ------------------------------------------------------------------------------ | Linearized educ2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- educ2 | status | 1.11e+07 5704049 1.95 0.051 -61287.73 2.23e+07 -------------+---------------------------------------------------------------- /cut_1_1 | -1.31e+08 2404031 -54.41 0.000 -1.36e+08 -1.26e+08 /cut_1_2 | -5818216 2957274 -1.97 0.049 -1.16e+07 -20465.59 /lnsig_1 | 18.72579 .0064207 2916.49 0.000 18.7132 18.73838 -------------+---------------------------------------------------------------- sig_1 | 1.36e+08 871138.6 1.34e+08 1.37e+08 ------------------------------------------------------------------------------
In both computers, this is the version of the command :
Code:
. which cmp /Users/celiavera/Library/Application Support/Stata/ado/plus/c/cmp.ado *! cmp 6.9.8 27 February 2016 *! Copyright (C) 2007-16 David Roodman *! Version history at bottom
If I remove the subpop() option or if I use $cmp_cont instead of $cmp_oprobit, the discrepancy disappears.
I appreciate very much your help.
Celia P.
Comment