--------------------------------------------------------------------------------------------
log: C:\AAA Miker Files\newer web pages\soc_388_notes\soc_388_2003\class 15.log
log type: text
opened on: 19 Nov 2003, 11:43:48
. set linesize 79
. use "C:\AAA Miker Files\newer web pages\soc_388_notes\Qian 80-90 intermar.dta
> ", clear
. *I'm going to use Qian's 80-90 intermarriage dataset, which has a lot of small
values and zeros, and I'm going to estimate a model that doesn't fit very well.
. desmat: poisson count mfulleth*med4*year ffulleth*fed4*year med4*fed4
------------------------------------------------------------------------------------------
Poisson regression
------------------------------------------------------------------------------------------
Dependent variable count
Optimization: ml
Number of observations: 512
Initial log likelihood: -1402408.286
Log likelihood: -159490.084
LR chi square: 2485836.405
Model degrees of freedom: 70
Pseudo R-squared: 0.886
Prob: 0.000
------------------------------------------------------------------------------------------
nr Effect Coeff s.e.
------------------------------------------------------------------------------------------
count
mfulleth
1 Hisp 5.733** 0.268
2 black 5.891** 0.268
3 white 7.874** 0.267
med4
4 2 0.787** 0.286
5 3 0.482 0.279
6 4 -0.797** 0.277
mfulleth.med4
7 Hisp.2 -1.657** 0.286
8 Hisp.3 -2.862** 0.280
9 Hisp.4 -4.208** 0.278
10 black.2 -1.144** 0.286
11 black.3 -2.383** 0.279
12 black.4 -3.577** 0.277
13 white.2 -0.765** 0.286
14 white.3 -1.903** 0.279
15 white.4 -2.367** 0.276
year
16 90 -0.511 0.579
mfulleth.year
17 Hisp.90 0.019 0.464
18 black.90 -0.285 0.464
19 white.90 0.286 0.463
med4.year
20 2.90 -0.446 0.504
21 3.90 -0.299 0.482
22 4.90 -0.419 0.477
mfulleth.med4.year
23 Hisp.2.90 0.492 0.505
24 Hisp.3.90 0.578 0.484
25 Hisp.4.90 0.155 0.480
26 black.2.90 0.584 0.505
27 black.3.90 0.672 0.483
28 black.4.90 0.095 0.479
29 white.2.90 0.256 0.504
30 white.3.90 0.267 0.482
31 white.4.90 -0.125 0.477
ffulleth
32 Hisp 5.344** 0.224
33 black 5.356** 0.224
34 white 7.489** 0.224
fed4
35 2 0.815** 0.238
36 3 0.142 0.235
37 4 -1.332** 0.236
ffulleth.fed4
38 Hisp.2 -1.591** 0.239
39 Hisp.3 -2.809** 0.236
40 Hisp.4 -4.194** 0.237
41 black.2 -1.032** 0.239
42 black.3 -1.982** 0.236
43 black.4 -3.067** 0.235
44 white.2 -0.660** 0.238
45 white.3 -1.700** 0.235
46 white.4 -2.285** 0.233
ffulleth.year
47 Hisp.90 -0.415 0.350
48 black.90 -0.650 0.350
49 white.90 -0.188 0.349
fed4.year
50 2.90 -1.002* 0.394
51 3.90 -0.352 0.372
52 4.90 0.177 0.364
ffulleth.fed4.year
53 Hisp.2.90 1.181** 0.395
54 Hisp.3.90 1.398** 0.374
55 Hisp.4.90 0.874* 0.369
56 black.2.90 1.094** 0.395
57 black.3.90 1.211** 0.373
58 black.4.90 0.522 0.367
59 white.2.90 0.899* 0.394
60 white.3.90 0.995** 0.372
61 white.4.90 0.402 0.365
med4.fed4
62 2.2 1.538** 0.010
63 2.3 1.794** 0.015
64 2.4 2.288** 0.035
65 3.2 1.893** 0.015
66 3.3 3.500** 0.018
67 3.4 4.422** 0.035
68 4.2 2.592** 0.034
69 4.3 4.786** 0.035
70 4.4 7.501** 0.046
71 _cons -5.887** 0.348
------------------------------------------------------------------------------------------
* p < .05
** p < .01
. poisgof
Goodness-of-fit chi2 = 316783.2
Prob > chi2(441) = 0.0000
. desmat: nbreg count mfulleth*med4*year ffulleth*fed4*year med4*fed4
------------------------------------------------------------------------------------------
Negative binomial regression
------------------------------------------------------------------------------------------
Dependent variable count
Optimization: ml
Number of observations: 512
Initial log likelihood: -2722.660
Log likelihood: -2417.521
LR chi square: 610.277
Model degrees of freedom: 70
Pseudo R-squared: 0.112
Dispersion: mean
Prob: 0.000
------------------------------------------------------------------------------------------
nr Effect Coeff s.e.
------------------------------------------------------------------------------------------
count
mfulleth
1 Hisp 5.136** 0.943
2 black 5.391** 1.058
3 white 5.930** 0.940
med4
4 2 2.109* 1.070
5 3 1.392 1.085
6 4 -0.145 1.109
mfulleth.med4
7 Hisp.2 -2.872* 1.143
8 Hisp.3 -3.548** 1.154
9 Hisp.4 -4.263** 1.169
10 black.2 -2.341* 1.142
11 black.3 -3.311** 1.152
12 black.4 -3.859** 1.168
13 white.2 -1.851 1.144
14 white.3 -2.689* 1.151
15 white.4 -2.633* 1.164
year
16 90 -2.565 1.457
mfulleth.year
17 Hisp.90 1.385 1.304
18 black.90 1.212 1.475
19 white.90 1.746 1.262
med4.year
20 2.90 0.652 1.434
21 3.90 1.274 1.405
22 4.90 1.573 1.428
mfulleth.med4.year
23 Hisp.2.90 -0.507 1.670
24 Hisp.3.90 -0.994 1.645
25 Hisp.4.90 -1.838 1.673
26 black.2.90 -0.573 1.667
27 black.3.90 -0.862 1.643
28 black.4.90 -1.558 1.670
29 white.2.90 -0.872 1.665
30 white.3.90 -1.171 1.637
31 white.4.90 -2.040 1.653
ffulleth
32 Hisp 4.450** 0.855
33 black 4.271** 1.023
34 white 5.435** 0.844
fed4
35 2 1.439 0.974
36 3 0.759 0.998
37 4 -0.558 1.023
ffulleth.fed4
38 Hisp.2 -1.950 1.042
39 Hisp.3 -3.153** 1.060
40 Hisp.4 -3.949** 1.085
41 black.2 -1.579 1.045
42 black.3 -2.339* 1.063
43 black.4 -2.837** 1.086
44 white.2 -1.125 1.045
45 white.3 -2.234* 1.059
46 white.4 -2.650* 1.081
ffulleth.year
47 Hisp.90 0.427 1.243
48 black.90 0.410 1.454
49 white.90 0.640 1.215
fed4.year
50 2.90 -1.494 1.292
51 3.90 0.004 1.297
52 4.90 0.038 1.340
ffulleth.fed4.year
53 Hisp.2.90 1.385 1.548
54 Hisp.3.90 0.807 1.554
55 Hisp.4.90 0.286 1.594
56 black.2.90 1.272 1.551
57 black.3.90 0.501 1.559
58 black.4.90 0.230 1.600
59 white.2.90 1.148 1.548
60 white.3.90 0.508 1.553
61 white.4.90 0.239 1.592
med4.fed4
62 2.2 1.438* 0.656
63 2.3 1.688* 0.657
64 2.4 1.612* 0.672
65 3.2 1.817** 0.649
66 3.3 3.409** 0.648
67 3.4 4.062** 0.671
68 4.2 2.186** 0.659
69 4.3 4.402** 0.664
70 4.4 6.830** 0.690
71 _cons -3.205** 1.050
lnalpha
72 _cons 1.042** 0.060
------------------------------------------------------------------------------------------
* p < .05
** p < .01
. *That's the same model, with nbreg instead of poisson. It has one extra term, the over-
> dispersion term lnalpha.
. *What's interesting here is how radically different the SE of the med4*fed4 terms are.
. If we get the direct nbreg output, which desmat initially suppresses.
unrecognized command: If
r(199);
. nbreg count _x_*
Fitting Poisson model:
Iteration 0: log likelihood = -2689974.6
Iteration 1: log likelihood = -2219812.2 (backed up)
Iteration 2: log likelihood = -1154089.9 (backed up)
Iteration 3: log likelihood = -534181.21
Iteration 4: log likelihood = -308895.05
Iteration 5: log likelihood = -161058.51
Iteration 6: log likelihood = -159495.85
Iteration 7: log likelihood = -159490.08
Iteration 8: log likelihood = -159490.08
Fitting constant-only model:
Iteration 0: log likelihood = -4060.4348
Iteration 1: log likelihood = -2731.4998
Iteration 2: log likelihood = -2722.6662
Iteration 3: log likelihood = -2722.6601
Iteration 4: log likelihood = -2722.6601
Fitting full model:
Iteration 0: log likelihood = -2668.6447 (not concave)
Iteration 1: log likelihood = -2592.5781 (not concave)
Iteration 2: log likelihood = -2512.3773
Iteration 3: log likelihood = -2472.7721
Iteration 4: log likelihood = -2427.693
Iteration 5: log likelihood = -2417.7676
Iteration 6: log likelihood = -2417.5218
Iteration 7: log likelihood = -2417.5213
Iteration 8: log likelihood = -2417.5213
Negative binomial regression Number of obs = 512
LR chi2(70) = 610.28
Prob > chi2 = 0.0000
Log likelihood = -2417.5213 Pseudo R2 = 0.1121
------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_x_1 | 5.135537 .9431738 5.44 0.000 3.28695 6.984124
_x_2 | 5.391223 1.057818 5.10 0.000 3.317939 7.464507
_x_3 | 5.929622 .9402622 6.31 0.000 4.086742 7.772502
_x_4 | 2.109165 1.070128 1.97 0.049 .0117521 4.206578
_x_5 | 1.392109 1.084754 1.28 0.199 -.7339689 3.518187
_x_6 | -.1452501 1.109393 -0.13 0.896 -2.319621 2.029121
_x_7 | -2.87222 1.143232 -2.51 0.012 -5.112913 -.6315275
_x_8 | -3.548247 1.154227 -3.07 0.002 -5.81049 -1.286004
_x_9 | -4.262664 1.169383 -3.65 0.000 -6.554614 -1.970715
_x_10 | -2.341135 1.142224 -2.05 0.040 -4.579854 -.1024162
_x_11 | -3.310937 1.151746 -2.87 0.004 -5.568317 -1.053557
_x_12 | -3.858745 1.167779 -3.30 0.001 -6.14755 -1.569939
_x_13 | -1.85068 1.144311 -1.62 0.106 -4.093489 .3921278
_x_14 | -2.688822 1.150703 -2.34 0.019 -4.944159 -.4334846
_x_15 | -2.632566 1.16385 -2.26 0.024 -4.913669 -.3514622
_x_16 | -2.565414 1.456543 -1.76 0.078 -5.420185 .2893569
_x_17 | 1.385326 1.304149 1.06 0.288 -1.17076 3.941411
_x_18 | 1.211602 1.474589 0.82 0.411 -1.67854 4.101743
_x_19 | 1.745791 1.262248 1.38 0.167 -.7281686 4.219751
_x_20 | .6519337 1.433718 0.45 0.649 -2.158102 3.46197
_x_21 | 1.274393 1.405239 0.91 0.364 -1.479825 4.028612
_x_22 | 1.573442 1.427934 1.10 0.271 -1.225257 4.372141
_x_23 | -.5071242 1.66993 -0.30 0.761 -3.780128 2.765879
_x_24 | -.9935789 1.645023 -0.60 0.546 -4.217764 2.230606
_x_25 | -1.838472 1.672759 -1.10 0.272 -5.117019 1.440075
_x_26 | -.5729003 1.66705 -0.34 0.731 -3.840258 2.694458
_x_27 | -.8624201 1.642769 -0.52 0.600 -4.082189 2.357348
_x_28 | -1.55807 1.670022 -0.93 0.351 -4.831253 1.715113
_x_29 | -.8718762 1.665196 -0.52 0.601 -4.1356 2.391847
_x_30 | -1.170809 1.637108 -0.72 0.475 -4.379482 2.037864
_x_31 | -2.04004 1.653416 -1.23 0.217 -5.280676 1.200595
_x_32 | 4.449884 .8545054 5.21 0.000 2.775084 6.124684
_x_33 | 4.270677 1.022905 4.18 0.000 2.26582 6.275534
_x_34 | 5.434985 .8443748 6.44 0.000 3.780041 7.089929
_x_35 | 1.438952 .9741681 1.48 0.140 -.4703819 3.348287
_x_36 | .7591769 .9983391 0.76 0.447 -1.197532 2.715886
_x_37 | -.5583202 1.023069 -0.55 0.585 -2.563499 1.446859
_x_38 | -1.949838 1.041544 -1.87 0.061 -3.991226 .0915502
_x_39 | -3.153348 1.059797 -2.98 0.003 -5.230511 -1.076185
_x_40 | -3.948891 1.085297 -3.64 0.000 -6.076034 -1.821748
_x_41 | -1.579013 1.045123 -1.51 0.131 -3.627416 .4693897
_x_42 | -2.339241 1.063418 -2.20 0.028 -4.423502 -.2549793
_x_43 | -2.837092 1.085973 -2.61 0.009 -4.965561 -.708623
_x_44 | -1.125296 1.045337 -1.08 0.282 -3.174118 .923527
_x_45 | -2.234005 1.059009 -2.11 0.035 -4.309625 -.1583849
_x_46 | -2.649929 1.081231 -2.45 0.014 -4.769103 -.5307557
_x_47 | .427068 1.243119 0.34 0.731 -2.009401 2.863537
_x_48 | .4102428 1.454459 0.28 0.778 -2.440444 3.26093
_x_49 | .640423 1.215052 0.53 0.598 -1.741035 3.021881
_x_50 | -1.493797 1.292274 -1.16 0.248 -4.026607 1.039014
_x_51 | .0039871 1.2967 0.00 0.998 -2.537499 2.545473
_x_52 | .0377585 1.339818 0.03 0.978 -2.588236 2.663753
_x_53 | 1.384957 1.547684 0.89 0.371 -1.648448 4.418361
_x_54 | .8073317 1.554281 0.52 0.603 -2.239002 3.853666
_x_55 | .2860036 1.594101 0.18 0.858 -2.838376 3.410383
_x_56 | 1.271529 1.551448 0.82 0.412 -1.769253 4.312312
_x_57 | .5010196 1.558793 0.32 0.748 -2.554158 3.556197
_x_58 | .2297772 1.600047 0.14 0.886 -2.906257 3.365811
_x_59 | 1.147584 1.547722 0.74 0.458 -1.885895 4.181062
_x_60 | .5076217 1.552895 0.33 0.744 -2.535997 3.551241
_x_61 | .2387873 1.59207 0.15 0.881 -2.881613 3.359187
_x_62 | 1.437679 .6563954 2.19 0.029 .1511677 2.72419
_x_63 | 1.687504 .656953 2.57 0.010 .3999003 2.975109
_x_64 | 1.61244 .6715112 2.40 0.016 .2963028 2.928578
_x_65 | 1.816864 .6488996 2.80 0.005 .5450438 3.088684
_x_66 | 3.409369 .6476067 5.26 0.000 2.140083 4.678655
_x_67 | 4.061697 .6708596 6.05 0.000 2.746836 5.376557
_x_68 | 2.185599 .6593864 3.31 0.001 .893225 3.477972
_x_69 | 4.401689 .6640965 6.63 0.000 3.100084 5.703294
_x_70 | 6.829921 .69005 9.90 0.000 5.477448 8.182394
_cons | -3.205368 1.049949 -3.05 0.002 -5.26323 -1.147506
-------------+----------------------------------------------------------------
/lnalpha | 1.041749 .0599969 .9241569 1.15934
-------------+----------------------------------------------------------------
alpha | 2.834169 .1700413 2.519743 3.18783
------------------------------------------------------------------------------
Likelihood-ratio test of alpha=0: chibar2(01) = 3.1e+05 Prob>=chibar2 = 0.000
. *We get a chisquare (here called chibar for a couple of technical reasons)
test of 310,000 or so.
. *log likelihood of the nbreg model is -2417
. *Log likelihood of the poisson model for the same variables is -159490
. display (-2*(-159490))-(-2*(-2417))
314146
. *the negative binomial version fits dramatically better here, and has some different things to say about educational intermarriage.
. *We have the difference of -2 log likelihood as 310,000
. exit, clear