Dear Statalist,
I am looking for help with my Cox regression (a survival analysis).
My unit of analysis is countries, the dependent variable is the adoption of a certain law (of course binary), and the time variable is years passed since 1990 to adoption year. My observation starts in 1990 and ends in 2005, each country is observed once per year.
My variables are time-varying ones (ln_ncbmn_cy ln_gdppercapita ln_population polity2 imfd igo_c_lltc5). On top of these time-varying variables, I included year dummies.
When I run a Cox model with the year dummies, it does not calculate standard errors for the dummies(a result attached below). It calculates SEs for everything correct, but not for year dummies.
Not only year dummies, but any variable that does not vary across countries within a year (for example, the number of countries that adopted the law in each year), it refuses to calculate SEs.
I ask statlist members if you have any advice on my problem.
Thanks,
David
I am looking for help with my Cox regression (a survival analysis).
My unit of analysis is countries, the dependent variable is the adoption of a certain law (of course binary), and the time variable is years passed since 1990 to adoption year. My observation starts in 1990 and ends in 2005, each country is observed once per year.
My variables are time-varying ones (ln_ncbmn_cy ln_gdppercapita ln_population polity2 imfd igo_c_lltc5). On top of these time-varying variables, I included year dummies.
When I run a Cox model with the year dummies, it does not calculate standard errors for the dummies(a result attached below). It calculates SEs for everything correct, but not for year dummies.
Not only year dummies, but any variable that does not vary across countries within a year (for example, the number of countries that adopted the law in each year), it refuses to calculate SEs.
I ask statlist members if you have any advice on my problem.
Thanks,
David
Code:
. stcox ln_ncbmn_cy ln_gdppercapita ln_population polity2 imfd igo_c_ll
> tc5 i.year
failure _d: ta1 == 1
analysis time _t: year_1
id: newid3
Iteration 0: log likelihood = -422.719
Iteration 1: log likelihood = -389.2535
Iteration 2: log likelihood = -387.77064
Iteration 3: log likelihood = -387.76954
Iteration 4: log likelihood = -387.76954 (backed up)
Iteration 5: log likelihood = -387.76954 (backed up)
Iteration 6: log likelihood = -387.76954 (backed up)
Refining estimates:
Iteration 0: log likelihood = -387.76954
Cox regression -- Breslow method for ties
No. of subjects = 151 Number of obs = 1,255
No. of failures = 93
Time at risk = 1255
LR chi2(6) = 69.90
Log likelihood = -387.76954 Prob > chi2 = 0.0000
---------------------------------------------------------------------------------
_t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
ln_ncbmn_cy | 1.02704 .0235104 1.17 0.244 .9819788 1.074169
ln_gdppercapita | 1.415395 .1743046 2.82 0.005 1.111868 1.801783
ln_population | 1.249535 .1106441 2.52 0.012 1.050452 1.486349
polity2 | 1.061197 .0209052 3.02 0.003 1.021004 1.102972
imfd | 1.103042 .3321092 0.33 0.745 .6113757 1.990105
igo_c_lltc5 | 1.022244 .5754177 0.04 0.969 .339169 3.081009
|
year |
1991 | 34.41589 . . . . .
1992 | 16.68557 . . . . .
1993 | 14.46724 . . . . .
1994 | 148.7342 . . . . .
1995 | 87.68588 . . . . .
1996 | 17.66053 . . . . .
1997 | 7.89e+07 . . . . .
1998 | 2.790774 . . . . .
1999 | 2.78696 . . . . .
2000 | 23.96774 . . . . .
2001 | .9907272 . . . . .
2002 | 2.790877 . . . . .
2003 | 5.950447 . . . . .
2004 | 12.60604 . . . . .
2005 | 8.103457 . . . . .
---------------------------------------------------------------------------------
.
end of do-file
Comment