Hello all.
I have trip data from a travel cost survey. It is overdispersed and truncated at zero, so I use ztnb to model it.
As it is collected on site, I have also run nbstrat which accounts for endogenous stratification. The estimates for alpha produced by nbstrat are extremely large. Can anyone explain to me why this is and if it signifies some kind of problem? I haven't been able to find the answer in the help document for this command.
Here is the code and the model results for ztnb and nbstrat:
I have trip data from a travel cost survey. It is overdispersed and truncated at zero, so I use ztnb to model it.
As it is collected on site, I have also run nbstrat which accounts for endogenous stratification. The estimates for alpha produced by nbstrat are extremely large. Can anyone explain to me why this is and if it signifies some kind of problem? I haven't been able to find the answer in the help document for this command.
Here is the code and the model results for ztnb and nbstrat:
HTML Code:
. ztnb trips tc_1 tcs dog_walking swimming age female full_time second_level wee > kend if trips > 0 & trips < 300 & distance < 31, nolog Zero-truncated negative binomial regression Number of obs = 218 LR chi2(9) = 69.50 Dispersion: mean Prob > chi2 = 0.0000 Log likelihood = -1069.6135 Pseudo R2 = 0.0315 ------------------------------------------------------------------------------ trips | Coefficient Std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- tc_1 | -.331683 .0470049 -7.06 0.000 -.423811 -.2395551 tcs | .0526015 .0288292 1.82 0.068 -.0039026 .1091056 dog_walking | .6912429 .1907299 3.62 0.000 .317419 1.065067 swimming | .6717875 .2134434 3.15 0.002 .2534461 1.090129 age | .0121518 .006755 1.80 0.072 -.0010878 .0253913 female | -.1296886 .16108 -0.81 0.421 -.4453997 .1860224 full_time | -.1954571 .2029611 -0.96 0.336 -.5932534 .2023393 second_level | .2632545 .1981498 1.33 0.184 -.1251119 .6516209 weekend | -.4113873 .1688693 -2.44 0.015 -.7423652 -.0804095 _cons | 3.877994 .4457129 8.70 0.000 3.004413 4.751576 -------------+---------------------------------------------------------------- /lnalpha | .2720226 .1195559 .0376974 .5063478 -------------+---------------------------------------------------------------- alpha | 1.312617 .156931 1.038417 1.65922 ------------------------------------------------------------------------------ Likelihood-ratio test of alpha=0: chibar2(01) = 1.1e+04 Prob>=chibar2 = 0.000 . nbstrat trips tc_1 tcs dog_walking swimming age female full_time second_level > weekend if trips > 0 & trips < 300 & distance < 31, nolog Negative Binomial with Endogenous Stratification Number of obs = 218 Wald chi2(9) = 126.13 Log likelihood = -1072.6005 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ trips | Coefficient Std. err. z P>|z| [95% conf. interval] -------------+---------------------------------------------------------------- tc_1 | -.3247509 .0417973 -7.77 0.000 -.4066722 -.2428296 tcs | .0501741 .0252458 1.99 0.047 .0006933 .0996549 dog_walking | .6690027 .1664985 4.02 0.000 .3426717 .9953338 swimming | .6589314 .1871537 3.52 0.000 .2921169 1.025746 age | .0119964 .005931 2.02 0.043 .0003717 .023621 female | -.1288407 .1417759 -0.91 0.363 -.4067164 .1490351 full_time | -.1942539 .178486 -1.09 0.276 -.54408 .1555723 second_level | .254796 .1740872 1.46 0.143 -.0864087 .5960007 weekend | -.4023662 .1483279 -2.71 0.007 -.6930836 -.1116488 _cons | -12.5788 143.0695 -0.09 0.930 -292.9899 267.8323 -------------+---------------------------------------------------------------- /lnalpha | 16.48533 143.0689 0.12 0.908 -263.9247 296.8953 -------------+---------------------------------------------------------------- alpha | 1.44e+07 2.07e+09 2.4e-115 8.7e+128 ------------------------------------------------------------------------------ AIC Statistic = 9.932 BIC Statistic = -1119.975 Deviance = 0.000 Dispersion = 0.000 .
