I am estimating a probit model with individual-level data on sickness and district-level data on soil contamination. The model estimates the probability of getting sick from soil contamination and I am estimating the marginal effects of the contamination. I have seven variables on contaminations such as arsenic, cadmium, mercury, etc. I have cross-sectional data at the district level. My model runs ok without district fixed effects and with additional control variables. However, when I add i.district (district fixed effects) to the model, many districts are omitted because of collinearity, and Stata says "convergence" not achieved after 16,000 iterations.
I would appreciate any help and hints on how to fix this issue.
my command: probit sick ave_s_arsenic ave_s_cadmium ave_s_copper ave_s_mercury ave_s_nickel ave_s_lead ave_s_zinc hhsize age female schooling married i.district, robust
foreach c in ave_s_arsenic ave_s_cadmium ave_s_copper ave_s_mercury ave_s_nickel ave_s_lead ave_s_zinc hhsize age female schooling married _cons { margins, dydx(`c') atmeans contrast }
Below is the output with the control variables and fixed effects.
Iteration 15999: log pseudolikelihood = | -897.2352 (not concave) | |
Iteration 16000: log pseudolikelihood = | -897.2352 (not concave) | |
convergence not achieved | ||
Probit regression | Number of obs = | 4,476 |
Wald chi2(27) = | 301.78 | |
Prob > chi2 = | 0.0000 | |
Log pseudolikelihood = -897.2352 | Pseudo R2 = | 0.1008 |
Robust | ||
sick Coef. Std. Err. | z P>z [95% Conf. | Interval] |
ave_s_arsenic .0151015 .127374 | 0.12 0.906 -.2345469 | .2647498 |
ave_s_cadmium .061447 .1980178 | 0.31 0.756 -.3266607 | .4495547 |
ave_s_copper .0911823 .1574646 | 0.58 0.563 -.2174426 | .3998073 |
ave_s_mercury -.0042525 .042269 | -0.10 0.920 -.0870983 | .0785932 |
ave_s_nickel -.0592163 .0252869 | -2.34 0.019 -.1087777 | -.0096549 |
ave_s_lead -.1195374 .2328372 | -0.51 0.608 -.5758899 | .336815 |
ave_s_zinc -.025928 .2773928 | -0.09 0.926 -.5696079 | .5177518 |
hhsize -.0220666 .022386 | -0.99 0.324 -.0659425 | .0218092 |
age .0147476 .0019474 | 7.57 0.000 .0109308 | .0185644 |
female .1434831 .0635808 | 2.26 0.024 .0188669 | .2680993 |
schooling -.0242093 .0069835 | -3.47 0.001 -.0378968 | -.0105219 |
married -.1073514 .0758817 | -1.41 0.157 -.2560769 | .041374 |
district | ||
Matad .4957028 .2886804 | 1.72 0.086 -.0701004 | 1.061506 |
Baruun-Urt .7779955 1.052935 | 0.74 0.460 -1.285719 | 2.84171 |
Sukhbaatar .7638064 .8018887 | 0.95 0.341 -.8078665 | 2.335479 |
Tuvshinshiree -.0674407 .3914831 | -0.17 0.863 -.8347335 | .6998521 |
Kherlen .6869447 .3192918 | 2.15 0.031 .0611442 | 1.312745 |
Batnorov .6739175 .296393 | 2.27 0.023 .0929979 | 1.254837 |
Bayan-Ovoo .4073151 .467568 | 0.87 0.384 -.5091014 | 1.323732 |
Darkhan .6915699 .3796485 | 1.82 0.069 -.0525276 | 1.435667 |
Tsenkhermandal .6399382 .3723019 | 1.72 0.086 -.0897602 | 1.369637 |
Bayan .8625859 .4420719 | 1.95 0.051 -.0038591 | 1.729031 |
Jargalant 1.53748 .8141955 | 1.89 0.059 -.0583141 | 3.133274 |
Zaamar .651218 .3193779 | 2.04 0.041 .0252488 | 1.277187 |
Sergelen .2986526 .4610042 | 0.65 0.517 -.604899 | 1.202204 |
Ugtaal 1.021395 .2698989 | 3.78 0.000 .4924027 | 1.550387 |
Erdene 0 (omitted) | ||
Bayangol 0 (omitted) | ||
Eroo 0 (omitted) | ||
Airag 0 (omitted) | ||
Dalanjargalan .4977266 .2686392 | 1.85 0.064 -.0287967 | 1.02425 |
Orgon 0 (omitted) | ||
Darkhan 0 (omitted) | ||
_cons -1.836731 . | . . . | . |
convergence not achieved | ||
r(430); | ||
end of do-file | ||
r(430); | ||
Odmaa
0 Response to Probit model with fixed effects
Post a Comment