-----------------------------------------------------------------------------------------------------
name: <unnamed>
log: C:\Users\mexmi\Documents\newer web pages\Soc_382\logs\2nd loglin intermar class.log
log type: text
opened on: 31 Jan 2019, 10:18:08
* Before class I was creating the crossings parameters, which I did not get to explain in class (and will next class).
. * replace cross1=1 if (meth_num==1 & feth_num>1) | (feth_num==1& meth_num>1)
. gen cross1=0
. replace cross1=1 if (meth_num==1 & feth_num>1) | (feth_num==1& meth_num>1)
(8 real changes made)
. gen cross2=0
. replace cross2=1 if (meth_num<=2 & feth_num>2)| (feth_num<=2 & meth_num>2)
(12 real changes made)
. gen cross3=0
. replace cross3=1 if (meth_num<=3 & feth_num>3)| (feth_num<=3 & meth_num>3)
(12 real changes made)
. gen cross4=0
. replace cross4=1 if (meth_num<=4 & feth_num>4)| (feth_num<=4 & meth_num>4)
(8 real changes made)
. poisson count i.meth_num i.feth_num i.cross*
Iteration 0: log likelihood = -1628492.6
Iteration 1: log likelihood = -514542.18 (backed up)
Iteration 2: log likelihood = -405521.33 (backed up)
Iteration 3: log likelihood = -241959.01 (backed up)
Iteration 4: log likelihood = -147645.83
Iteration 5: log likelihood = -59259.102
Iteration 6: log likelihood = -12308.206
Iteration 7: log likelihood = -10257.604
Iteration 8: log likelihood = -10246.31
Iteration 9: log likelihood = -10246.309
Poisson regression Number of obs = 25
LR chi2(12) = 3132470.69
Prob > chi2 = 0.0000
Log likelihood = -10246.309 Pseudo R2 = 0.9935
-------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------------+----------------------------------------------------------------
meth_num |
Mexican American | -.977803 .0181359 -53.92 0.000 -1.013349 -.9422572
Hispanic Other | -1.938943 .020284 -95.59 0.000 -1.978698 -1.899187
Non Hispanic Other | -2.506451 .0216409 -115.82 0.000 -2.548866 -2.464035
White non Hispanic | .6765227 .0185343 36.50 0.000 .6401961 .7128493
|
feth_num |
Mexican American | -.0199891 .018195 -1.10 0.272 -.0556507 .0156725
Hispanic Other | -.8181779 .0202753 -40.35 0.000 -.8579168 -.778439
Non Hispanic Other | -1.107257 .0209967 -52.73 0.000 -1.14841 -1.066104
White non Hispanic | 1.871097 .0185343 100.95 0.000 1.83477 1.907423
|
1.cross1 | -3.026666 .017803 -170.01 0.000 -3.061559 -2.991773
1.cross2 | -1.014698 .0101682 -99.79 0.000 -1.034627 -.9947684
1.cross3 | -.5411649 .0113079 -47.86 0.000 -.563328 -.5190018
1.cross4 | -1.444942 .0092927 -155.49 0.000 -1.463155 -1.426728
_cons | 10.65775 .0048495 2197.69 0.000 10.64825 10.66726
-------------------------------------------------------------------------------------
. poisgof
Deviance goodness-of-fit = 20262.62
Prob > chi2(12) = 0.0000
Pearson goodness-of-fit = 21598.27
Prob > chi2(12) = 0.0000
* We started class here, talking about where the Quasi-independence model fit the data, and where it didn’t.
. table meth_num feth_num, contents(sum count sum quasi_indep_model) row col cellwidth(10)
--------------------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non Total
--------------------+-----------------------------------------------------------------------
Black, non Hispanic | 42521 291 412 393 2064 45681
| 42521 274.1622 229.8974 264.6786 2391.262 45681
|
Mexican American | 94 18088 612 433 6067 25294
| 108.2118 18088 565.4383 650.9836 5881.366 25294
|
Hispanic Other | 310 633 5901 258 4507 11609
| 84.44069 526.1821 5901 507.9809 4589.396 11609
|
Non Hispanic Other | 101 317 214 3509 3959 8100
| 68.72013 428.2213 359.0829 3509 3734.976 8100
|
White non Hispanic | 615 5338 4403 5505 543276 559137
| 858.6274 5350.435 4486.582 5165.357 543276 559137
|
Total | 43641 24667 11542 10098 559873 649821
| 43641 24667 11542 10098 559873 649821
--------------------------------------------------------------------------------------------
* I generated the cell-by-cell pearson chisquare statistic, in order to figure out which cells fit the data worst by the quasi-symmetry model. Since the Pearson statistic is positive everywhere, it is more useful on a cell-by-cell basis than the LR chisquare statistic, which is positive in some cells and negative in others.
. gen pearson_quasi_indep=((count- quasi_indep_model)^2)/ quasi_indep_model
. table meth_num feth_num, contents(sum count sum quasi_indep_model sum pearson_quasi_indep ) row col cellwidth(10)
--------------------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non Total
--------------------+-----------------------------------------------------------------------
Black, non Hispanic | 42521 291 412 393 2064 45681
| 42521 274.1622 229.8974 264.6786 2391.262 45681
| 0 1.034097 144.2441 62.21268 44.78817 252.279
|
Mexican American | 94 18088 612 433 6067 25294
| 108.2118 18088 565.4383 650.9836 5881.366 25294
| 1.86647 0 3.834181 72.99238 5.859167 84.5522
|
Hispanic Other | 310 633 5901 258 4507 11609
| 84.44069 526.1821 5901 507.9809 4589.396 11609
| 602.5176 21.68462 0 123.0173 1.479319 748.6989
|
Non Hispanic Other | 101 317 214 3509 3959 8100
| 68.72013 428.2213 359.0829 3509 3734.976 8100
| 15.1628 28.88735 58.61894 0 13.43702 116.1061
|
White non Hispanic | 615 5338 4403 5505 543276 559137
| 858.6274 5350.435 4486.582 5165.357 543276 559137
| 69.12699 .0288983 1.55706 22.3329 0 93.04585
|
Total | 43641 24667 11542 10098 559873 649821
| 43641 24667 11542 10098 559873 649821
| 688.6739 51.63497 208.2543 280.5553 65.56367 1294.682
--------------------------------------------------------------------------------------------
* Looking at this table, we see that cells 1,3 and 3,1 (black-other Hispanic intermarriage) contribute a lot to the total Pearson chisquare statistic of 1294.
. codebook meth_num
-----------------------------------------------------------------------------------------------------
meth_num husband's race/ethnicity
-----------------------------------------------------------------------------------------------------
type: numeric (byte)
label: ethnicity
range: [1,5] units: 1
unique values: 5 missing .: 0/25
tabulation: Freq. Numeric Label
5 1 Black, non Hispanic
5 2 Mexican American
5 3 Hispanic Other
5 4 Non Hispanic Other
5 5 White non Hispanic
* Now I generate the symmetric dummy variable to fit the two black-other Hispanic cells.
. gen byte black_othHisp=0
. replace black_othHisp=1 if (meth_num==1 & feth_num==3)|(feth_num==1 & meth_num==3)
(2 real changes made)
. *this is the gender symmetric black-Other Hispanic interaction.
. *Which we picked because the black-other Hispanic cells are fit poorly by quasi-symmetry
. poisson count i.meth_num i.feth_num i.endogamy_diagonal_cat
Iteration 0: log likelihood = -1622248.5
Iteration 1: log likelihood = -451561.88 (backed up)
Iteration 2: log likelihood = -374791.98 (backed up)
Iteration 3: log likelihood = -199026.43 (backed up)
Iteration 4: log likelihood = -143118.6
Iteration 5: log likelihood = -64630.014
Iteration 6: log likelihood = -2062.1525
Iteration 7: log likelihood = -650.23659
Iteration 8: log likelihood = -648.54413
Iteration 9: log likelihood = -648.54411
Poisson regression Number of obs = 25
LR chi2(13) = 3151666.22
Prob > chi2 = 0.0000
Log likelihood = -648.54411 Pseudo R2 = 0.9996
---------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------+----------------------------------------------------------------
meth_num |
Mexican American | .899968 .0213908 42.07 0.000 .8580427 .9418932
Hispanic Other | .6519274 .0222121 29.35 0.000 .6083924 .6954623
Non Hispanic Other | .4459202 .0231596 19.25 0.000 .4005282 .4913121
White non Hispanic | 2.971213 .0235674 126.07 0.000 2.925022 3.017404
|
feth_num |
Mexican American | 1.829598 .032363 56.53 0.000 1.766168 1.893028
Hispanic Other | 1.653511 .0327399 50.50 0.000 1.589342 1.71768
Non Hispanic Other | 1.794394 .0323429 55.48 0.000 1.731004 1.857785
White non Hispanic | 3.995454 .0336881 118.60 0.000 3.929427 4.061482
|
endogamy_diagonal_cat |
1 | 6.873631 .0370393 185.58 0.000 6.801036 6.946227
2 | 3.289316 .0229991 143.02 0.000 3.244239 3.334393
3 | 2.593316 .0262895 98.64 0.000 2.54179 2.644843
4 | 2.13865 .0286843 74.56 0.000 2.08243 2.19487
5 | 2.454583 .0192917 127.24 0.000 2.416772 2.492394
|
_cons | 3.784122 .0367204 103.05 0.000 3.712151 3.856093
---------------------------------------------------------------------------------------
. poisgof
Deviance goodness-of-fit = 1067.088
Prob > chi2(11) = 0.0000
Pearson goodness-of-fit = 1294.682
Prob > chi2(11) = 0.0000
. *That was quasi independenc
. poisson count i.meth_num i.feth_num i.endogamy_diagonal_cat i.black_othHisp
Iteration 0: log likelihood = -1622150.1
Iteration 1: log likelihood = -448544.31 (backed up)
Iteration 2: log likelihood = -406174.34 (backed up)
Iteration 3: log likelihood = -164842.09 (backed up)
Iteration 4: log likelihood = -122114.39 (backed up)
Iteration 5: log likelihood = -38616.448
Iteration 6: log likelihood = -5553.1389
Iteration 7: log likelihood = -541.68591
Iteration 8: log likelihood = -407.86486
Iteration 9: log likelihood = -407.61852
Iteration 10: log likelihood = -407.61852
Poisson regression Number of obs = 25
LR chi2(14) = 3152148.07
Prob > chi2 = 0.0000
Log likelihood = -407.61852 Pseudo R2 = 0.9997
---------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------+----------------------------------------------------------------
meth_num |
Mexican American | 1.011966 .022452 45.07 0.000 .9679614 1.055972
Hispanic Other | .7375092 .0229166 32.18 0.000 .6925934 .782425
Non Hispanic Other | .5583559 .0241463 23.12 0.000 .51103 .6056819
White non Hispanic | 3.178408 .0259865 122.31 0.000 3.127476 3.229341
|
feth_num |
Mexican American | 1.952846 .0332485 58.73 0.000 1.88768 2.018012
Hispanic Other | 1.713306 .0330711 51.81 0.000 1.648488 1.778124
Non Hispanic Other | 1.920179 .0332427 57.76 0.000 1.855024 1.985333
White non Hispanic | 4.213423 .0356215 118.28 0.000 4.143607 4.28324
|
endogamy_diagonal_cat |
1 | 7.180968 .0406743 176.55 0.000 7.101248 7.260688
2 | 3.361407 .0238147 141.15 0.000 3.314731 3.408082
3 | 2.755277 .0277498 99.29 0.000 2.700889 2.809666
4 | 2.207766 .0293102 75.32 0.000 2.150319 2.265213
5 | 2.336756 .0206074 113.39 0.000 2.296366 2.377146
|
1.black_othHisp | 1.072106 .0447908 23.94 0.000 .9843172 1.159894
_cons | 3.476785 .0403842 86.09 0.000 3.397634 3.555937
---------------------------------------------------------------------------------------
* Among black married people who are not married to other black people, the log odds of being married to a person from Other-Hispanic background was significantly higher (increased the log odds by 1.07, increased the count by a factor of exp(1.07), or had a relative risk of exp(1.07)) compared to marriages to whites, non-Hispanics, and Mexican Americans. That is, among the non endogamous black married people, more than expected were married to people from the other Hispanic category.
. poisgof
Deviance goodness-of-fit = 585.237
Prob > chi2(10) = 0.0000
Pearson goodness-of-fit = 602.9136
Prob > chi2(10) = 0.0000
* Adds one df, for black-other Hispanic interaction, and improves the fit by (1067-585=482) on 1 df, which is highly significant.
. predict QI_and_BOH
(option n assumed; predicted number of events)
. table meth_num feth_num, contents(sum count sum QI_and_BOH) row col cellwidth(10)
--------------------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non Total
--------------------+-----------------------------------------------------------------------
Black, non Hispanic | 42521 291 412 393 2064 45681
| 42521 228.0651 524.3694 220.7353 2186.83 45681
|
Mexican American | 94 18088 612 433 6067 25294
| 89.01025 18088 493.7638 607.2439 6015.982 25294
|
Hispanic Other | 310 633 5901 258 4507 11609
| 197.6306 476.8205 5901 461.4959 4572.053 11609
|
Non Hispanic Other | 101 317 214 3509 3959 8100
| 56.5509 398.6114 313.703 3509 3822.135 8100
|
White non Hispanic | 615 5338 4403 5505 543276 559137
| 776.8082 5475.503 4309.164 5299.525 543276 559137
|
Total | 43641 24667 11542 10098 559873 649821
| 43641 24667 11542 10098 559873 649821
--------------------------------------------------------------------------------------------
. display 310+412
722
. display 197.63+524.37
722
*We have one term fitting the two cells for black-Other Hispanic intermarriage, so the sum of those two cells is the same in the actual and the predicted.
* Now add the gender specific black-other Hispanic term.
. gen byte husb_black_wife_OH=0
. replace husb_black_wife_OH=1 if meth_num==1 & feth_num==3
(1 real change made)
. table meth_num feth_num, contents(mean black_othHisp) cellwidth(10)
note: cellwidth too small, variable name truncated;
to increase cellwidth, specify cellwidth(#)
--------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non
--------------------+-----------------------------------------------------------
Black, non Hispanic | 0 0 1 0 0
Mexican American | 0 0 0 0 0
Hispanic Other | 1 0 0 0 0
Non Hispanic Other | 0 0 0 0 0
White non Hispanic | 0 0 0 0 0
--------------------------------------------------------------------------------
. table meth_num feth_num, contents(mean husb_black_wife_OH ) cellwidth(10)
note: cellwidth too small, variable name truncated;
to increase cellwidth, specify cellwidth(#)
--------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non
--------------------+-----------------------------------------------------------
Black, non Hispanic | 0 0 1 0 0
Mexican American | 0 0 0 0 0
Hispanic Other | 0 0 0 0 0
Non Hispanic Other | 0 0 0 0 0
White non Hispanic | 0 0 0 0 0
--------------------------------------------------------------------------------
. poisson count i.meth_num i.feth_num i.endogamy_diagonal_cat i.black_othHisp i.husb_black_wife_OH
Iteration 0: log likelihood = -1622147.3
Iteration 1: log likelihood = -448422.28 (backed up)
Iteration 2: log likelihood = -405890.17 (backed up)
Iteration 3: log likelihood = -164210.39 (backed up)
Iteration 4: log likelihood = -121629.82 (backed up)
Iteration 5: log likelihood = -38857.644
Iteration 6: log likelihood = -5529.6282
Iteration 7: log likelihood = -491.60417
Iteration 8: log likelihood = -355.79505
Iteration 9: log likelihood = -355.54662
Iteration 10: log likelihood = -355.54662
Poisson regression Number of obs = 25
LR chi2(15) = 3152252.22
Prob > chi2 = 0.0000
Log likelihood = -355.54662 Pseudo R2 = 0.9998
---------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------+----------------------------------------------------------------
meth_num |
Mexican American | .9702156 .0224704 43.18 0.000 .9261743 1.014257
Hispanic Other | .6751551 .0234341 28.81 0.000 .6292252 .721085
Non Hispanic Other | .5166051 .0241635 21.38 0.000 .4692455 .5639646
White non Hispanic | 3.136658 .0260024 120.63 0.000 3.085694 3.187621
|
feth_num |
Mexican American | 2.082757 .0372736 55.88 0.000 2.009703 2.155812
Hispanic Other | 1.864941 .0377599 49.39 0.000 1.790933 1.938949
Non Hispanic Other | 2.05009 .0372684 55.01 0.000 1.977046 2.123135
White non Hispanic | 4.343335 .0394049 110.22 0.000 4.266103 4.420567
|
endogamy_diagonal_cat |
1 | 7.269129 .0429614 169.20 0.000 7.184926 7.353332
2 | 3.361407 .0238147 141.15 0.000 3.314731 3.408082
3 | 2.754156 .0277426 99.28 0.000 2.699782 2.808531
4 | 2.207766 .0293102 75.32 0.000 2.150319 2.265213
5 | 2.336756 .0206074 113.39 0.000 2.296366 2.377146
|
1.black_othHisp | 1.672793 .0697805 23.97 0.000 1.536025 1.80956
1.husb_black_wife_OH | -.9053349 .0873382 -10.37 0.000 -1.076515 -.7341552
_cons | 3.388624 .0426868 79.38 0.000 3.30496 3.472289
---------------------------------------------------------------------------------------
. poisgof
Deviance goodness-of-fit = 481.0932
Prob > chi2(9) = 0.0000
Pearson goodness-of-fit = 494.695
Prob > chi2(9) = 0.0000
. *Now let's see what black-other Hispanic intermarriage looks like of we don't mark out the endogamy
> diagonal.
. poisson count i.meth_num i.feth_num i.black_othHisp
Iteration 0: log likelihood = -312574.59
Iteration 1: log likelihood = -232297.17
Iteration 2: log likelihood = -224728.61
Iteration 3: log likelihood = -224682.36
Iteration 4: log likelihood = -224682.36
Poisson regression Number of obs = 25
LR chi2(9) = 2703598.60
Prob > chi2 = 0.0000
Log likelihood = -224682.36 Pseudo R2 = 0.8575
-------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------------+----------------------------------------------------------------
meth_num |
Mexican American | -.6016685 .0078446 -76.70 0.000 -.6170436 -.5862934
Hispanic Other | -1.341195 .0104305 -128.58 0.000 -1.361639 -1.320752
Non Hispanic Other | -1.740372 .0120606 -144.30 0.000 -1.76401 -1.716733
White non Hispanic | 2.494159 .0048776 511.35 0.000 2.484599 2.503719
|
feth_num |
Mexican American | -.5811256 .0079728 -72.89 0.000 -.5967521 -.5654992
Hispanic Other | -1.299468 .010508 -123.66 0.000 -1.320063 -1.278872
Non Hispanic Other | -1.474255 .0110479 -133.44 0.000 -1.495908 -1.452601
White non Hispanic | 2.541118 .0049812 510.15 0.000 2.531355 2.550881
|
1.black_othHisp | -.8394343 .0379587 -22.11 0.000 -.9138319 -.7650366
_cons | 8.048426 .0066011 1219.26 0.000 8.035488 8.061364
-------------------------------------------------------------------------------------
*Although the black-Other Hispanic association was positive and significant compared to other forms of non endogamy for black spouses, that is compared to marriage to other non-black groups, when we get rid of the terms for the endogamy diagonal, the black-other Hispanic association is strongly negative, because now we are also comparing to black-black marriage, and that is always more common. Also, without the endogamy diagonal, the model will fit poorly.
. poisgof
Deviance goodness-of-fit = 449134.7
Prob > chi2(15) = 0.0000
Pearson goodness-of-fit = 1153768
Prob > chi2(15) = 0.0000
. *when not controlling for the endogamy diagonal, models tend to fit poorly
. gen QS=0
. replace QS= (meth_num*10+ feth_num) if feth_num>meth_num
(10 real changes made)
. replace QS= (feth_num*10+ meth_num) if meth_num>feth_num
(10 real changes made)
*Constructing a set of numbers that are symmetric and otherwise different in every cell.
. table meth_num feth_num, contents(mean QS) cellwidth(10)
--------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non
--------------------+-----------------------------------------------------------
Black, non Hispanic | 0 12 13 14 15
Mexican American | 12 0 23 24 25
Hispanic Other | 13 23 0 34 35
Non Hispanic Other | 14 24 34 0 45
White non Hispanic | 15 25 35 45 0
--------------------------------------------------------------------------------
* Quasi symmetry treats the terms as
. poisson count i.meth_num i.feth_num i.QS
Iteration 0: log likelihood = -1621737
Iteration 1: log likelihood = -438727.68 (backed up)
Iteration 2: log likelihood = -371941.3 (backed up)
Iteration 3: log likelihood = -207096.11 (backed up)
Iteration 4: log likelihood = -185110.43
Iteration 5: log likelihood = -90707.459
Iteration 6: log likelihood = -16934.19
Iteration 7: log likelihood = -573.01331
Iteration 8: log likelihood = -173.51256
Iteration 9: log likelihood = -170.2886
Iteration 10: log likelihood = -170.28779
Iteration 11: log likelihood = -170.28779
Poisson regression Number of obs = 25
LR chi2(18) = 3152622.73
Prob > chi2 = 0.0000
Log likelihood = -170.28779 Pseudo R2 = 0.9999
-------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------------+----------------------------------------------------------------
meth_num |
Mexican American | -.8865297 .0195915 -45.25 0.000 -.9249284 -.8481311
Hispanic Other | -1.464572 .0203211 -72.07 0.000 -1.504401 -1.424743
Non Hispanic Other | -1.916449 .0213965 -89.57 0.000 -1.958386 -1.874513
White non Hispanic | .7607015 .0177297 42.91 0.000 .7259519 .7954511
|
feth_num |
Mexican American | .0317804 .0195915 1.62 0.105 -.0066182 .0701791
Hispanic Other | -.5103041 .0203211 -25.11 0.000 -.5501328 -.4704754
Non Hispanic Other | -.5782177 .0213965 -27.02 0.000 -.6201542 -.5362813
White non Hispanic | 1.786918 .0177297 100.79 0.000 1.752168 1.821668
|
QS |
12 | -5.072186 .0518097 -97.90 0.000 -5.173731 -4.970641
13 | -3.891192 .0387965 -100.30 0.000 -3.967232 -3.815152
14 | -4.109943 .047238 -87.01 0.000 -4.202528 -4.017358
15 | -4.857751 .021174 -229.42 0.000 -4.899251 -4.81625
23 | -2.809359 .0293163 -95.83 0.000 -2.866818 -2.7519
24 | -3.078001 .037751 -81.53 0.000 -3.151991 -3.00401
25 | -2.856983 .0101084 -282.64 0.000 -2.876795 -2.837171
34 | -2.977466 .0473135 -62.93 0.000 -3.070199 -2.884733
35 | -2.54299 .012457 -204.14 0.000 -2.567405 -2.518574
45 | -2.234246 .0134023 -166.71 0.000 -2.260515 -2.207978
|
_cons | 10.65775 .0048495 2197.69 0.000 10.64825 10.66726
-------------------------------------------------------------------------------------
. poisgof
Deviance goodness-of-fit = 110.5756
Prob > chi2(6) = 0.0000
Pearson goodness-of-fit = 117.1273
Prob > chi2(6) = 0.0000
. poisson count i.meth_num i.feth_num i.endogamy_diagonal_cat i.QS
* Even though it is not obvious why this must be so, the QS terms are the full set of symmetric interaction terms that can be included, so adding the endogamy diagonal terms (which are also symmetric with respect to husband and wife) yield the same model, but with a different set of terms displayed:
note: 15.QS omitted because of collinearity
note: 25.QS omitted because of collinearity
note: 34.QS omitted because of collinearity
note: 35.QS omitted because of collinearity
note: 45.QS omitted because of collinearity
Iteration 0: log likelihood = -1621737
Iteration 1: log likelihood = -438727.66 (backed up)
Iteration 2: log likelihood = -371941.28 (backed up)
Iteration 3: log likelihood = -207096.16 (backed up)
Iteration 4: log likelihood = -185110.46
Iteration 5: log likelihood = -90707.453
Iteration 6: log likelihood = -16934.185
Iteration 7: log likelihood = -573.01339
Iteration 8: log likelihood = -173.51256
Iteration 9: log likelihood = -170.2886
Iteration 10: log likelihood = -170.28779
Iteration 11: log likelihood = -170.28779
Poisson regression Number of obs = 25
LR chi2(18) = 3152622.73
Prob > chi2 = 0.0000
Log likelihood = -170.28779 Pseudo R2 = 0.9999
---------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------+----------------------------------------------------------------
meth_num |
Mexican American | 1.114238 .0245816 45.33 0.000 1.066059 1.162417
Hispanic Other | .850189 .025271 33.64 0.000 .8006588 .8997193
Non Hispanic Other | .707055 .0260209 27.17 0.000 .6560549 .7580551
White non Hispanic | 3.818683 .0528711 72.23 0.000 3.715057 3.922308
|
feth_num |
Mexican American | 2.032548 .0343926 59.10 0.000 1.96514 2.099956
Hispanic Other | 1.804457 .0346164 52.13 0.000 1.73661 1.872304
Non Hispanic Other | 2.045286 .0344563 59.36 0.000 1.977753 2.11282
White non Hispanic | 4.844899 .058194 83.25 0.000 4.730841 4.958957
|
endogamy_diagonal_cat |
1 | 7.915732 .0643011 123.10 0.000 7.789704 8.04176
2 | 3.914197 .0524013 74.70 0.000 3.811492 4.016901
3 | 3.286209 .0501069 65.58 0.000 3.188002 3.384417
4 | 2.668723 .0513256 52.00 0.000 2.568127 2.76932
5 | 1.79977 .0483798 37.20 0.000 1.704947 1.894592
|
QS |
12 | .8427781 .0735992 11.45 0.000 .6985263 .98703
13 | 1.709778 .0634045 26.97 0.000 1.585508 1.834049
14 | 1.182285 .0681776 17.34 0.000 1.048659 1.31591
15 | 0 (omitted)
23 | .7908445 .0558517 14.16 0.000 .6813773 .9003117
24 | .2134592 .0604599 3.53 0.000 .0949599 .3319584
25 | 0 (omitted)
34 | 0 (omitted)
35 | 0 (omitted)
45 | 0 (omitted)
|
_cons | 2.742022 .064118 42.77 0.000 2.616353 2.867691
---------------------------------------------------------------------------------------
. poisgof
Deviance goodness-of-fit = 110.5756
Prob > chi2(6) = 0.0000
Pearson goodness-of-fit = 117.1273
Prob > chi2(6) = 0.0000
. *This is another version of the QS, quasi symmetry model, which has the endogamy diagonal displa
> cing 5 of the off-diagonal QS terms.
. *gen ID_QS=(50/649821)*(abs(count- constant_only_class ))
. predict QS_predicted
(option n assumed; predicted number of events)
. *gen ID_QS=(50/649821)*(abs(count- QS_predicted ))
. gen ID_QS=(50/649821)*(abs(count- QS_predicted ))
. table meth_num feth_num, contents(sum ID_QS) row col cellwidth(10)
--------------------------------------------------------------------------------------------
husband's | wife's race/ethnicity
race/ethnicity | Black, non Mexican Am Hispanic O Non Hispan White non Total
--------------------+-----------------------------------------------------------------------
Black, non Hispanic | 0 .0012189 .0084073 .0001272 .0070611 .0168145
Mexican American | .0012189 0 .001669 .001508 .0043959 .0087917
Hispanic Other | .0084073 .001669 0 .0017512 .008325 .0201525
Non Hispanic Other | .0001272 .001508 .0017512 0 .0031319 .0065183
White non Hispanic | .0070611 .0043959 .008325 .0031319 0 .0229139
|
Total | .0168145 .0087917 .0201525 .0065183 .0229139 .075191
--------------------------------------------------------------------------------------------
. poisson count i.meth_num i.feth_num i.endogamy_diagonal_cat i.black_othHisp i.husb_black_wife_OH
Iteration 0: log likelihood = -1622147.3
Iteration 1: log likelihood = -448422.28 (backed up)
Iteration 2: log likelihood = -405890.17 (backed up)
Iteration 3: log likelihood = -164210.39 (backed up)
Iteration 4: log likelihood = -121629.82 (backed up)
Iteration 5: log likelihood = -38857.644
Iteration 6: log likelihood = -5529.6282
Iteration 7: log likelihood = -491.60417
Iteration 8: log likelihood = -355.79505
Iteration 9: log likelihood = -355.54662
Iteration 10: log likelihood = -355.54662
Poisson regression Number of obs = 25
LR chi2(15) = 3152252.22
Prob > chi2 = 0.0000
Log likelihood = -355.54662 Pseudo R2 = 0.9998
---------------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------+----------------------------------------------------------------
meth_num |
Mexican American | .9702156 .0224704 43.18 0.000 .9261743 1.014257
Hispanic Other | .6751551 .0234341 28.81 0.000 .6292252 .721085
Non Hispanic Other | .5166051 .0241635 21.38 0.000 .4692455 .5639646
White non Hispanic | 3.136658 .0260024 120.63 0.000 3.085694 3.187621
|
feth_num |
Mexican American | 2.082757 .0372736 55.88 0.000 2.009703 2.155812
Hispanic Other | 1.864941 .0377599 49.39 0.000 1.790933 1.938949
Non Hispanic Other | 2.05009 .0372684 55.01 0.000 1.977046 2.123135
White non Hispanic | 4.343335 .0394049 110.22 0.000 4.266103 4.420567
|
endogamy_diagonal_cat |
1 | 7.269129 .0429614 169.20 0.000 7.184926 7.353332
2 | 3.361407 .0238147 141.15 0.000 3.314731 3.408082
3 | 2.754156 .0277426 99.28 0.000 2.699782 2.808531
4 | 2.207766 .0293102 75.32 0.000 2.150319 2.265213
5 | 2.336756 .0206074 113.39 0.000 2.296366 2.377146
|
1.black_othHisp | 1.672793 .0697805 23.97 0.000 1.536025 1.80956
1.husb_black_wife_OH | -.9053349 .0873382 -10.37 0.000 -1.076515 -.7341552
_cons | 3.388624 .0426868 79.38 0.000 3.30496 3.472289
---------------------------------------------------------------------------------------
. * lincom can test whether two terms, for instance two endogamy terms, are significantly different from each other. To test whether the force of endogamy is the same across groups or different, that would be a goodness of fit test between the two models, one which had a single term for endogamy, and the Quasi-Independence model with 5 endogamy terms.
.
. lincom 4.endogamy_diagonal_cat-5.endogamy_diagonal_cat
( 1) [count]4.endogamy_diagonal_cat - [count]5.endogamy_diagonal_cat = 0
------------------------------------------------------------------------------
count | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | -.1289895 .0431923 -2.99 0.003 -.2136449 -.0443342
------------------------------------------------------------------------------
*How do we know that the goodness of fit test at the top of the model is compared to the constant only model? We run the constant only model and compute the -2 times log likelihood of the difference of the two models.
. display (-1576481-(-355.54))*-2
3152250.9
. log close
name: <unnamed>
log: C:\Users\mexmi\Documents\newer web pages\Soc_382\logs\2nd loglin intermar class.log
log type: text
closed on: 31 Jan 2019, 14:04:47
--------------------------------------------------------------------------------------------------