------------------------------------------------------------------------------------------------------ log: c:\Imbook\bwebpage\Section4\mma18p1heterogeneity.txt log type: text opened on: 19 May 2005, 17:58:22 . . ********** OVERVIEW OF MMA18P1HETEROGENEITY.DO ********** . . * STATA Program . * copyright C 2005 by A. Colin Cameron and Pravin K. Trivedi . * used for "Microeconometrics: Methods and Applications" . * by A. Colin Cameron and Pravin K. Trivedi (2005) . * Cambridge University Press . . * Chapter 18.8 Pages 632-6 . * Unobserved Heterogeneity with Duration data Example . * (1) Exponential with and without heterogeneity . * Residuals Plots: Figures 18.2 (exp.wmf) and 18.3 (exp_gamma.wmf) . * Tabulate Model Estimates: Table 18.1 . * (2) Weibull with and without heterogeneity: Generalized Residuals Plots . * Residuals Plots: Figures 18.4 (Weibul16.wmf) and 18.5 (Weibul16_IG.wmf) . * Tabulate model Estimates: Table 18.2 . . * To run this program you need data file . * ema1996.dta . . ********** SETUP ********** . . set more off . version 8.0 . set scheme s1mono /* Used for graphs */ . set matsize 100 . . ********** DATA DESCRIPTION ********** . . * The data is from . * B.P. McCall (1996), "Unemployment Insurance Rules, Joblessness, . * and Part-time Work," Econometrica, 64, 647-682. . . * There are 3343 observations from the CPS Displaced Worker Surveys . * of 1986, 1988, 1990 and 1992 on 33 variables including . * spell = length of spell in number of two-week intervals . * CENSOR1 = 1 if re-employed at full-time job . . * See program mma17p4duration.do for further description of the data set . . ********** READ DATA ********** . . use ema1996.dta (Sample for 1996 EMA paper: part-time= worked part-time last week) . . ********** CREATE ADDITIONAL VARIABLES ********** . . gen RR = reprate . gen DR = disrate . gen UI = ui . gen RRUI = RR*UI . gen DRUI = DR*UI . gen LOGWAGE = logwage . sum Variable | Obs Mean Std. Dev. Min Max -------------+-------------------------------------------------------- spell | 3343 6.247981 5.611271 1 28 censor1 | 3343 .3209692 .4669188 0 1 censor2 | 3343 .1014059 .3019106 0 1 censor3 | 3343 .1717021 .3771777 0 1 censor4 | 3343 .3754113 .4843014 0 1 -------------+-------------------------------------------------------- ui | 3343 .5527969 .4972791 0 1 reprate | 3343 .4544717 .1137918 .066 2.059 logwage | 3343 5.692994 .5356591 2.70805 7.600402 tenure | 3343 4.114867 5.862322 0 40 disrate | 3343 .1094376 .0735274 .002 1.02 -------------+-------------------------------------------------------- slack | 3343 .4884834 .4999421 0 1 abolpos | 3343 .1456775 .3528354 0 1 explose | 3343 .5025426 .5000683 0 1 stateur | 3343 6.5516 1.803825 2.5 13 houshead | 3343 .6120251 .4873617 0 1 -------------+-------------------------------------------------------- married | 3343 .5860006 .4926221 0 1 female | 3343 .3478911 .4763725 0 1 child | 3343 .4501944 .4975876 0 1 ychild | 3343 .1956327 .3967463 0 1 nonwhite | 3343 .1390966 .3460991 0 1 -------------+-------------------------------------------------------- age | 3343 35.44331 10.6402 20 61 schlt12 | 3343 .2811846 .4496446 0 1 schgt12 | 3343 .3356267 .4722797 0 1 smsa | 3343 .7241998 .4469835 0 1 bluecoll | 3343 .6036494 .489212 0 1 -------------+-------------------------------------------------------- mining | 3343 .029315 .1687132 0 1 constr | 3343 .1480706 .3552231 0 1 transp | 3343 .0646126 .2458778 0 1 trade | 3343 .1848639 .3882452 0 1 fire | 3343 .0514508 .2209484 0 1 -------------+-------------------------------------------------------- services | 3343 .1699073 .3756075 0 1 pubadmin | 3343 .0095722 .097383 0 1 year85 | 3343 .2677236 .442839 0 1 year87 | 3343 .2174693 .4125862 0 1 year89 | 3343 .1998205 .3999251 0 1 -------------+-------------------------------------------------------- midatl | 3343 .1088842 .3115405 0 1 encen | 3343 .1429853 .3501103 0 1 wncen | 3343 .0643135 .2453472 0 1 southatl | 3343 .2375112 .4256217 0 1 escen | 3343 .0532456 .2245564 0 1 -------------+-------------------------------------------------------- wscen | 3343 .1441819 .3513266 0 1 mountain | 3343 .1079868 .3104102 0 1 pacific | 3343 .0260245 .159232 0 1 RR | 3343 .4544717 .1137918 .066 2.059 DR | 3343 .1094376 .0735274 .002 1.02 -------------+-------------------------------------------------------- UI | 3343 .5527969 .4972791 0 1 RRUI | 3343 .2478687 .2380667 0 2.059 DRUI | 3343 .0602776 .0754261 0 .824 LOGWAGE | 3343 5.692994 .5356591 2.70805 7.600402 . . ********* ANALYSIS: UNEMPLOYMENT DURATION ********** . . * Stata st curves require defining the dependent variable . * and the censoring variable if there is one . stset spell, fail(censor1=1) failure event: censor1 == 1 obs. time interval: (0, spell] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 20887 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 28 . stdes failure _d: censor1 == 1 analysis time _t: spell |-------------- per subject --------------| Category total mean min median max ------------------------------------------------------------------------------ no. of subjects 3343 no. of records 3343 1 1 1 1 (first) entry time 0 0 0 0 (final) exit time 6.247981 1 5 28 subjects with gap 0 time on gap if gap 0 time at risk 20887 6.247981 1 5 28 failures 1073 .3209692 0 0 1 ------------------------------------------------------------------------------ . . * Define $xlist = list of regressors used in subsequent regressions . global xlist RR DR UI RRUI DRUI LOGWAGE /* > */ tenure slack abolpos explose stateur houshead married /* > */ female child ychild nonwhite age schlt12 schgt12 smsa bluecoll /* > */ mining constr transp trade fire services pubadmin /* > */ year85 year87 year89 midatl /* > */ encen wncen southatl escen wscen mountain pacific . . * (1) EXPONENTIAL REGRESSION . . * Estimate exponential without heterogeneity . streg $xlist, nolog nohr dist(exponential) robust failure _d: censor1 == 1 analysis time _t: spell Exponential regression -- log relative-hazard form No. of subjects = 3343 Number of obs = 3343 No. of failures = 1073 Time at risk = 20887 Wald chi2(40) = 565.24 Log pseudo-likelihood = -2700.6903 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- RR | .4720235 .6005534 0.79 0.432 -.7050396 1.649087 DR | -.5756396 .7624489 -0.75 0.450 -2.070012 .9187327 UI | -1.424561 .2493917 -5.71 0.000 -1.91336 -.9357622 RRUI | .9655904 .6118408 1.58 0.115 -.2335956 2.164776 DRUI | -.1990635 1.019118 -0.20 0.845 -2.196498 1.798371 LOGWAGE | .3508005 .115598 3.03 0.002 .1242327 .5773684 tenure | -.0001462 .0064637 -0.02 0.982 -.0128147 .0125224 slack | -.2593666 .0759363 -3.42 0.001 -.4081991 -.1105342 abolpos | -.1550897 .0953306 -1.63 0.104 -.3419342 .0317549 explose | .198458 .0648354 3.06 0.002 .071383 .3255331 stateur | -.064626 .0229903 -2.81 0.005 -.1096862 -.0195659 houshead | .3812208 .0836602 4.56 0.000 .2172499 .5451918 married | .369552 .0786145 4.70 0.000 .2154705 .5236335 female | .1164067 .0852986 1.36 0.172 -.0507754 .2835888 child | -.0333008 .0794577 -0.42 0.675 -.1890352 .1224335 ychild | -.1449722 .1022781 -1.42 0.156 -.3454336 .0554892 nonwhite | -.6692066 .1188272 -5.63 0.000 -.9021037 -.4363095 age | -.0220821 .0039256 -5.63 0.000 -.0297762 -.0143879 schlt12 | -.1231414 .0966102 -1.27 0.202 -.3124939 .066211 schgt12 | .1114395 .082945 1.34 0.179 -.0511297 .2740087 smsa | .1922291 .0799904 2.40 0.016 .0354508 .3490075 bluecoll | -.2033718 .085129 -2.39 0.017 -.3702215 -.036522 mining | -.1205818 .1973575 -0.61 0.541 -.5073955 .2662319 constr | -.04475 .1081519 -0.41 0.679 -.2567237 .1672238 transp | -.1786694 .156034 -1.15 0.252 -.4844906 .1271517 trade | -.0345159 .1019152 -0.34 0.735 -.234266 .1652341 fire | .1120549 .1386716 0.81 0.419 -.1597365 .3838462 services | .1840002 .0983911 1.87 0.061 -.0088428 .3768432 pubadmin | .1090606 .2954211 0.37 0.712 -.4699541 .6880752 year85 | .2147661 .0888664 2.42 0.016 .0405911 .388941 year87 | .3541162 .0948499 3.73 0.000 .1682139 .5400186 year89 | .467082 .1104355 4.23 0.000 .2506325 .6835316 midatl | .0264112 .1465647 0.18 0.857 -.2608503 .3136727 encen | .0043916 .1502813 0.03 0.977 -.2901544 .2989375 wncen | .1724311 .1607689 1.07 0.283 -.1426703 .4875324 southatl | .2638807 .1183726 2.23 0.026 .0318747 .4958867 escen | .35414 .19317 1.83 0.067 -.0244664 .7327463 wscen | .3385896 .1433308 2.36 0.018 .0576664 .6195128 mountain | .0063693 .1538821 0.04 0.967 -.2952341 .3079727 pacific | .0770202 .2393505 0.32 0.748 -.3920982 .5461385 _cons | -4.079107 .8767097 -4.65 0.000 -5.797426 -2.360788 ------------------------------------------------------------------------------ . estimates store bexp . . * Figure 18.2 (p.633) - Generalized (Cox-Snell) Residuals for Exponential . predict resid, csnell . stset resid, fail(censor1) failure event: censor1 != 0 & censor1 < . obs. time interval: (0, resid] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 1073 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 5.218098 . sts generate survivor=s . generate cumhaz = -ln(survivor) . sort resid . graph twoway (scatter cumhaz resid, c(J) msymbol(i) msize(small) clstyle(p1)) /* > */ (scatter resid resid, c(l) msymbol(i) msize(small) clstyle(p2)), /* > */ scale (1.2) plotregion(style(none)) /* > */ title("Exponential Model Residuals") /* > */ xtitle("Generalized (Cox-Snell) Residual", size(medlarge)) xscale(titlegap(*5)) /* > */ ytitle("Cumulative Hazard", size(medlarge)) yscale(titlegap(*5)) /* > */ legend(pos(6) ring(0) col(1)) legend(size(small)) /* > */ legend( label(1 "Cumulative Hazard") label(2 "45 degree line")) . graph export exp.wmf, replace (file c:\Imbook\bwebpage\Section4\exp.wmf written in Windows Metafile format) . drop resid survivor cumhaz . . * Estimate exponential with gamma heterogeneity . stset spell, fail(censor1) failure event: censor1 != 0 & censor1 < . obs. time interval: (0, spell] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 20887 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 28 . streg $xlist, nolog nohr dist(exponential) frailty(gamma) robust failure _d: censor1 analysis time _t: spell Exponential regression -- log relative-hazard form Gamma frailty No. of subjects = 3343 Number of obs = 3343 No. of failures = 1073 Time at risk = 20887 Wald chi2(40) = 576.86 Log pseudo-likelihood = -2695.3518 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- RR | .5005828 .6187508 0.81 0.419 -.7121465 1.713312 DR | -.8824469 .7894395 -1.12 0.264 -2.42972 .664826 UI | -1.584537 .2622252 -6.04 0.000 -2.098489 -1.070586 RRUI | 1.091168 .6327026 1.72 0.085 -.1489067 2.331242 DRUI | .0574048 1.047123 0.05 0.956 -1.994919 2.109729 LOGWAGE | .3792805 .1191278 3.18 0.001 .1457944 .6127666 tenure | .0007938 .0065903 0.12 0.904 -.012123 .0137106 slack | -.2862928 .0770348 -3.72 0.000 -.4372782 -.1353074 abolpos | -.1842749 .0977213 -1.89 0.059 -.3758051 .0072552 explose | .2151452 .0663117 3.24 0.001 .0851767 .3451137 stateur | -.0650451 .023552 -2.76 0.006 -.1112061 -.0188841 houshead | .3960399 .0847153 4.67 0.000 .2300009 .5620789 married | .3961194 .0806744 4.91 0.000 .2380005 .5542384 female | .1102564 .0869256 1.27 0.205 -.0601147 .2806275 child | -.0464355 .0815869 -0.57 0.569 -.206343 .113472 ychild | -.1213622 .103309 -1.17 0.240 -.3238441 .0811196 nonwhite | -.6909793 .1217489 -5.68 0.000 -.9296027 -.4523559 age | -.0225342 .0040184 -5.61 0.000 -.0304101 -.0146582 schlt12 | -.1513782 .0968026 -1.56 0.118 -.3411079 .0383515 schgt12 | .1011742 .0834622 1.21 0.225 -.0624088 .2647572 smsa | .212363 .081774 2.60 0.009 .052089 .372637 bluecoll | -.220439 .0862751 -2.56 0.011 -.3895351 -.0513429 mining | -.1721823 .2051663 -0.84 0.401 -.5743008 .2299362 constr | -.0897602 .11034 -0.81 0.416 -.3060225 .1265022 transp | -.1572488 .1563607 -1.01 0.315 -.4637102 .1492126 trade | -.0451107 .1034986 -0.44 0.663 -.2479642 .1577428 fire | .0881685 .1386688 0.64 0.525 -.1836175 .3599544 services | .1682835 .1005405 1.67 0.094 -.0287723 .3653393 pubadmin | .0961407 .3092103 0.31 0.756 -.5099004 .7021817 year85 | .1940199 .0906564 2.14 0.032 .0163366 .3717031 year87 | .3564373 .0959014 3.72 0.000 .1684741 .5444005 year89 | .4924007 .1101907 4.47 0.000 .2764308 .7083705 midatl | .0156736 .1488094 0.11 0.916 -.2759874 .3073347 encen | .0089345 .1538505 0.06 0.954 -.2926069 .3104759 wncen | .1742124 .1634726 1.07 0.287 -.1461881 .4946129 southatl | .2676635 .1192515 2.24 0.025 .0339348 .5013922 escen | .3741169 .199389 1.88 0.061 -.0166783 .7649121 wscen | .361461 .1423856 2.54 0.011 .0823903 .6405316 mountain | -.00019 .1557385 -0.00 0.999 -.3054318 .3050519 pacific | .0800478 .2463547 0.32 0.745 -.4027986 .5628941 _cons | -4.095067 .9086039 -4.51 0.000 -5.875898 -2.314236 -------------+---------------------------------------------------------------- /ln_the | -1.462995 .31608 -4.63 0.000 -2.0825 -.8434894 -------------+---------------------------------------------------------------- theta | .2315418 .0731857 .1246183 .4302067 ------------------------------------------------------------------------------ . estimates store bexpgamma . . * Figure 18.3 (p.633) - Generalized (Cox-Snell) Residuals for Exponential-Gamma . predict resid, csnell (option unconditional assumed) . stset resid, fail(censor1) failure event: censor1 != 0 & censor1 < . obs. time interval: (0, resid] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 1073 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 3.971096 . sts generate survivor=s . generate cumhaz = -ln(survivor) . sort resid . graph twoway (scatter cumhaz resid, c(J) msymbol(i) msize(small) clstyle(p1)) /* > */ (scatter resid resid, c(l) msymbol(i) msize(small) clstyle(p2)), /* > */ scale (1.2) plotregion(style(none)) /* > */ title("Exponential-Gamma Model Residuals") /* > */ xtitle("Generalized (Cox-Snell) Residual", size(medlarge)) xscale(titlegap(*5)) /* > */ ytitle("Cumulative Hazard", size(medlarge)) yscale(titlegap(*5)) /* > */ legend(pos(6) ring(0) col(1)) legend(size(small)) /* > */ legend( label(1 "Cumulative Hazard") label(2 "45 degree line")) . graph export exp_gamma.wmf, replace (file c:\Imbook\bwebpage\Section4\exp_gamma.wmf written in Windows Metafile format) . drop resid survivor cumhaz . . /* > * Following did not work, even with starting values provided > * Results in book obtained on different computer with different Stata version > * Estimate exponential with IG heterogeneity > stset spell, fail(censor1=1) > quietly streg $xlist, nolog nohr dist(exponential) robust > matrix theta = 1.6 > matrix bstart = e(b),theta > streg $xlist, nohr dist(exponential) frailty(invgauss) robust from(bstart) > * estimates store bexpIG > */ . . * Table 18.1 (p.634) - Display Parameter Estimates . * Note that exponetial-IG missing . estimates table bexp bexpgamma, t(%9.3f) stats(N ll) b(%9.3f) /* > */ keep(RR DR UI RRUI DRUI LOGWAGE _cons) -------------------------------------- Variable | bexp bexpgamma -------------+------------------------ RR | 0.472 0.501 | 0.786 0.809 DR | -0.576 -0.882 | -0.755 -1.118 UI | -1.425 -1.585 | -5.712 -6.043 RRUI | 0.966 1.091 | 1.578 1.725 DRUI | -0.199 0.057 | -0.195 0.055 LOGWAGE | 0.351 0.379 | 3.035 3.184 _cons | -4.079 -4.095 | -4.653 -4.507 -------------+------------------------ N | 3343.000 3343.000 ll | -2700.690 -2695.352 -------------------------------------- legend: b/t . . * (2) WEIBULL REGRESSION . . * Estimate Weibull without heterogeneity . stset spell, fail(censor1=1) failure event: censor1 == 1 obs. time interval: (0, spell] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 20887 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 28 . streg $xlist, nolog nohr dist(weibull) robust failure _d: censor1 == 1 analysis time _t: spell Weibull regression -- log relative-hazard form No. of subjects = 3343 Number of obs = 3343 No. of failures = 1073 Time at risk = 20887 Wald chi2(40) = 501.65 Log pseudo-likelihood = -2687.5995 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- RR | .4481156 .6381895 0.70 0.483 -.8027127 1.698944 DR | -.4269187 .8086983 -0.53 0.598 -2.011938 1.158101 UI | -1.496066 .2639679 -5.67 0.000 -2.013434 -.9786984 RRUI | 1.015226 .6455611 1.57 0.116 -.2500501 2.280503 DRUI | -.2988417 1.065384 -0.28 0.779 -2.386956 1.789272 LOGWAGE | .3655253 .12212 2.99 0.003 .1261745 .6048761 tenure | -.0011127 .0068716 -0.16 0.871 -.0145809 .0123554 slack | -.2652154 .0803214 -3.30 0.001 -.4226424 -.1077883 abolpos | -.1604227 .1012942 -1.58 0.113 -.3589557 .0381103 explose | .2075085 .0684715 3.03 0.002 .0733068 .3417103 stateur | -.0708745 .0242117 -2.93 0.003 -.1183286 -.0234204 houshead | .3976626 .0887192 4.48 0.000 .2237762 .571549 married | .3786057 .0830317 4.56 0.000 .2158665 .541345 female | .1260829 .0896987 1.41 0.160 -.0497233 .301889 child | -.0336778 .0839956 -0.40 0.688 -.1983061 .1309505 ychild | -.1613066 .108947 -1.48 0.139 -.3748389 .0522256 nonwhite | -.7025504 .12426 -5.65 0.000 -.9460956 -.4590052 age | -.0235823 .0041922 -5.63 0.000 -.0317989 -.0153658 schlt12 | -.1226759 .1022762 -1.20 0.230 -.3231335 .0777816 schgt12 | .1162848 .0880692 1.32 0.187 -.0563278 .2888973 smsa | .1999567 .0841129 2.38 0.017 .0350985 .3648149 bluecoll | -.1994925 .0899354 -2.22 0.027 -.3757626 -.0232223 mining | -.1015676 .2036644 -0.50 0.618 -.5007425 .2976073 constr | -.0253737 .1135609 -0.22 0.823 -.247949 .1972016 transp | -.1981522 .1672141 -1.19 0.236 -.5258858 .1295814 trade | -.0311361 .1079502 -0.29 0.773 -.2427146 .1804423 fire | .1262153 .1492527 0.85 0.398 -.1663145 .4187452 services | .2031673 .1038945 1.96 0.051 -.0004622 .4067968 pubadmin | .1117728 .3087374 0.36 0.717 -.4933415 .716887 year85 | .2374972 .093387 2.54 0.011 .054462 .4205325 year87 | .3787397 .1011782 3.74 0.000 .1804341 .5770454 year89 | .4920278 .1180472 4.17 0.000 .2606596 .7233959 midatl | .02465 .1542139 0.16 0.873 -.2776037 .3269036 encen | -.0014111 .1579065 -0.01 0.993 -.3109023 .30808 wncen | .1844363 .1694444 1.09 0.276 -.1476687 .5165413 southatl | .2740974 .1250481 2.19 0.028 .0290076 .5191872 escen | .367742 .2024771 1.82 0.069 -.0291058 .7645899 wscen | .3440005 .1527804 2.25 0.024 .0445563 .6434446 mountain | .0159627 .1620188 0.10 0.922 -.3015883 .3335136 pacific | .0849532 .2504077 0.34 0.734 -.4058368 .5757432 _cons | -4.357886 .9196792 -4.74 0.000 -6.160424 -2.555347 -------------+---------------------------------------------------------------- /ln_p | .1215314 .0194374 6.25 0.000 .0834348 .1596281 -------------+---------------------------------------------------------------- p | 1.129225 .0219492 1.087014 1.173075 1/p | .8855632 .0172131 .8524608 .9199511 ------------------------------------------------------------------------------ . estimates store bweib . . * Figure 18.4 (p.635) - Generalized (Cox-Snell) Residuals for Weibull . predict resid, csnell . stset resid, fail(censor1) failure event: censor1 != 0 & censor1 < . obs. time interval: (0, resid] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 1073 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 6.283261 . sts generate survivor=s . generate cumhaz = -ln(survivor) . sort resid . graph twoway (scatter cumhaz resid, c(J) msymbol(i) msize(small) clstyle(p1)) /* > */ (scatter resid resid, c(l) msymbol(i) msize(small) clstyle(p2)), /* > */ scale (1.2) plotregion(style(none)) /* > */ title("Weibull Model Residuals") /* > */ xtitle("Generalized (Cox-Snell) Residual", size(medlarge)) xscale(titlegap(*5)) /* > */ ytitle("Cumulative Hazard", size(medlarge)) yscale(titlegap(*5)) /* > */ legend(pos(6) ring(0) col(1)) legend(size(small)) /* > */ legend( label(1 "Cumulative Hazard") label(2 "45 degree line")) . graph export Weibul16.wmf, replace (file c:\Imbook\bwebpage\Section4\Weibul16.wmf written in Windows Metafile format) . drop resid survivor cumhaz . . * Estimate Weibull with gamma heterogeneity . stset spell, fail(censor1=1) failure event: censor1 == 1 obs. time interval: (0, spell] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 20887 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 28 . streg $xlist, nolog nohr dist(weibull) frailty(invgauss) robust failure _d: censor1 == 1 analysis time _t: spell Weibull regression -- log relative-hazard form Inverse-Gaussian frailty No. of subjects = 3343 Number of obs = 3343 No. of failures = 1073 Time at risk = 20887 Wald chi2(40) = 643.00 Log pseudo-likelihood = -2616.3216 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- RR | .7356277 .9058181 0.81 0.417 -1.039743 2.510998 DR | -1.072566 1.149098 -0.93 0.351 -3.324758 1.179625 UI | -2.574752 .3843798 -6.70 0.000 -3.328123 -1.821381 RRUI | 1.733571 .9333928 1.86 0.063 -.0958458 3.562987 DRUI | -.060621 1.537813 -0.04 0.969 -3.07468 2.953438 LOGWAGE | .575656 .1766599 3.26 0.001 .2294089 .9219031 tenure | -.0009848 .0097472 -0.10 0.920 -.0200889 .0181194 slack | -.4416007 .1142976 -3.86 0.000 -.6656199 -.2175814 abolpos | -.2873066 .1465357 -1.96 0.050 -.5745113 -.0001019 explose | .3641943 .0976897 3.73 0.000 .1727259 .5556627 stateur | -.0981133 .0346763 -2.83 0.005 -.1660775 -.030149 houshead | .5924383 .1256739 4.71 0.000 .3461219 .8387546 married | .6083214 .1183487 5.14 0.000 .3763624 .8402805 female | .1788439 .1285074 1.39 0.164 -.0730259 .4307137 child | -.0914227 .121778 -0.75 0.453 -.3301031 .1472578 ychild | -.1805373 .1527477 -1.18 0.237 -.4799173 .1188426 nonwhite | -1.008517 .1725174 -5.85 0.000 -1.346645 -.6703894 age | -.0333776 .0059183 -5.64 0.000 -.0449772 -.0217779 schlt12 | -.2258621 .1439543 -1.57 0.117 -.5080075 .0562832 schgt12 | .1505129 .124469 1.21 0.227 -.0934418 .3944677 smsa | .3009952 .119907 2.51 0.012 .0659819 .5360086 bluecoll | -.3211857 .1253163 -2.56 0.010 -.5668012 -.0755702 mining | -.2319827 .3008491 -0.77 0.441 -.8216361 .3576708 constr | -.1260324 .1633669 -0.77 0.440 -.4462257 .1941609 transp | -.2763858 .225893 -1.22 0.221 -.7191279 .1663562 trade | -.0687616 .1518284 -0.45 0.651 -.3663399 .2288166 fire | .0668973 .2131814 0.31 0.754 -.3509306 .4847252 services | .231914 .1494712 1.55 0.121 -.0610441 .5248721 pubadmin | .0901949 .4579252 0.20 0.844 -.807322 .9877117 year85 | .2780139 .1339053 2.08 0.038 .0155644 .5404634 year87 | .5208783 .1415375 3.68 0.000 .2434699 .7982867 year89 | .7209598 .1655487 4.35 0.000 .3964903 1.045429 midatl | -.0192077 .2222646 -0.09 0.931 -.4548382 .4164228 encen | -.0297055 .2284931 -0.13 0.897 -.4775438 .4181328 wncen | .2460338 .24216 1.02 0.310 -.2285911 .7206586 southatl | .3563643 .1793284 1.99 0.047 .0048872 .7078415 escen | .5461543 .2910193 1.88 0.061 -.024233 1.116542 wscen | .4606814 .2140966 2.15 0.031 .0410598 .880303 mountain | .017581 .2293804 0.08 0.939 -.4319963 .4671584 pacific | .1379886 .3636985 0.38 0.704 -.5748475 .8508247 _cons | -5.303059 1.34133 -3.95 0.000 -7.932017 -2.6741 -------------+---------------------------------------------------------------- /ln_p | .5611667 .0225898 24.84 0.000 .5168915 .6054418 /ln_the | 1.852696 .0896755 20.66 0.000 1.676935 2.028457 -------------+---------------------------------------------------------------- p | 1.752716 .0395935 1.676807 1.832062 1/p | .570543 .0128884 .5458332 .5963715 theta | 6.376987 .5718595 5.349136 7.602343 ------------------------------------------------------------------------------ . estimates store bweibIG . . * Figure 18.5 (p.636) - Generalized (Cox-Snell) Residuals for Weibull-IG . predict resid, csnell (option unconditional assumed) . stset resid, fail(censor1) failure event: censor1 != 0 & censor1 < . obs. time interval: (0, resid] exit on or before: failure ------------------------------------------------------------------------------ 3343 total obs. 0 exclusions ------------------------------------------------------------------------------ 3343 obs. remaining, representing 1073 failures in single record/single failure data 1073 total analysis time at risk, at risk from t = 0 earliest observed entry t = 0 last observed exit t = 5.044588 . sts generate survivor=s . generate cumhaz = -ln(survivor) . sort resid . graph twoway (scatter cumhaz resid, c(J) msymbol(i) msize(small) clstyle(p1)) /* > */ (scatter resid resid, c(l) msymbol(i) msize(small) clstyle(p2)), /* > */ scale (1.2) plotregion(style(none)) /* > */ title("Weibull-IG Model Residuals") /* > */ xtitle("Generalized (Cox-Snell) Residual", size(medlarge)) xscale(titlegap(*5)) /* > */ ytitle("Cumulative Hazard", size(medlarge)) yscale(titlegap(*5)) /* > */ legend(pos(6) ring(0) col(1)) legend(size(small)) /* > */ legend( label(1 "Cumulative Hazard") label(2 "45 degree line")) . graph export Weibul16_IG.wmf, replace (file c:\Imbook\bwebpage\Section4\Weibul16_IG.wmf written in Windows Metafile format) . drop resid survivor cumhaz . . * Table 18.2 (p.635) - Display Parameter Estimates . estimates table bweibIG bweib, t(%9.3f) stats(N ll) b(%9.3f) /* > */ keep(RR DR UI RRUI DRUI LOGWAGE _cons) -------------------------------------- Variable | bweibIG bweib -------------+------------------------ RR | 0.736 0.448 | 0.812 0.702 DR | -1.073 -0.427 | -0.933 -0.528 UI | -2.575 -1.496 | -6.698 -5.668 RRUI | 1.734 1.015 | 1.857 1.573 DRUI | -0.061 -0.299 | -0.039 -0.281 LOGWAGE | 0.576 0.366 | 3.259 2.993 _cons | -5.303 -4.358 | -3.954 -4.738 -------------+------------------------ N | 3343.000 3343.000 ll | -2616.322 -2687.600 -------------------------------------- legend: b/t . . ********** CLOSE OUTPUT ********** . log close log: c:\Imbook\bwebpage\Section4\mma18p1heterogeneity.txt log type: text closed on: 19 May 2005, 17:58:38 ----------------------------------------------------------------------------------------------------