---------------------------------------------------------------------------------------- log: D:\Anands Documents\Working Papers\Devanbu\Data\commit\analysis_fse.txt log type: text opened on: 19 Mar 2007, 16:56:28 . . use commit_apache . . /* > All models account for unobserved heterogeneity, i.e. they cluster observations > by id. > > The time pieces in all models are as follows > > tp1: 0-6 months > tp2: >6 months <= 1year > tp3: >1 year <= 2 years > tp4: >2 years <= 3 years > tp5: >3 years <= 4 years > tp6: >4 years > > The duration dependence results for Apache and Postgres show the same pattern -- it fi > rst increases and the decreases. > The rate increases rapidly upto tp2 (just before the end of 1 year) for Apache and upt > o tp3 (just before the end > of 2 years) for Postgres and then declines. The results show no evidence of duration > dependence in the rate of > being allowed commit priviliges in the Python project. > > Z-scores greater than 1.96 imply that the effect is statistically significant within a > 95% > confidence interval (the typical measure of significance for a 2-tailed test). > > patch_submitted is a count of patches submitted in that month > */ . . *The following two models are based on the Apache data . . stpiece time_trend devs_cum sent_cum patches_submitted norm_indegree_cum, tp(0) cluste > r(id) nohr Invoking stsplit... Creating time pieces... failure _d: commit == 1 analysis time _t: tf id: id Iteration 0: log pseudolikelihood = -195.56437 Iteration 1: log pseudolikelihood = -173.92187 Iteration 2: log pseudolikelihood = -130.06127 Iteration 3: log pseudolikelihood = -100.08253 Iteration 4: log pseudolikelihood = -96.878741 Iteration 5: log pseudolikelihood = -96.861305 Iteration 6: log pseudolikelihood = -96.861298 Exponential regression -- log relative-hazard form No. of subjects = 1445 Number of obs = 50170 No. of failures = 30 Time at risk = 4180.88849 Wald chi2(6) = 698.60 Log pseudolikelihood = -96.861298 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1445 clusters in id) ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- tp1 | -4.0249 1.640223 -2.45 0.014 -7.239678 -.8101216 time_trend | -.8661608 .4310416 -2.01 0.044 -1.710987 -.0213348 devs_cum | .1027477 .0631071 1.63 0.103 -.0209399 .2264353 sent_cum | .0196848 .0042045 4.68 0.000 .0114442 .0279255 patches_su~d | .0517945 .0308955 1.68 0.094 -.0087597 .1123486 norm_indeg~m | 52.51973 16.16963 3.25 0.001 20.82783 84.21162 ------------------------------------------------------------------------------ . stpiece time_trend devs_cum sent_cum patches_submitted norm_indegree_cum, tp(0,0.5,1,2 > ,3,4) cluster(id) nohr Invoking stsplit... Creating time pieces... failure _d: commit == 1 analysis time _t: tf id: id Iteration 0: log pseudolikelihood = -195.56437 Iteration 1: log pseudolikelihood = -173.63465 Iteration 2: log pseudolikelihood = -128.53202 Iteration 3: log pseudolikelihood = -95.547715 Iteration 4: log pseudolikelihood = -90.902727 Iteration 5: log pseudolikelihood = -90.7677 Iteration 6: log pseudolikelihood = -90.76746 Iteration 7: log pseudolikelihood = -90.76746 Exponential regression -- log relative-hazard form No. of subjects = 1445 Number of obs = 54823 No. of failures = 30 Time at risk = 4180.88849 Wald chi2(11) = 782.69 Log pseudolikelihood = -90.76746 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1445 clusters in id) ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- tp1 | -6.049215 1.953232 -3.10 0.002 -9.877479 -2.220952 tp2 | -5.404121 1.973235 -2.74 0.006 -9.271591 -1.53665 tp3 | -5.890668 2.084012 -2.83 0.005 -9.975256 -1.80608 tp4 | -8.042755 2.403216 -3.35 0.001 -12.75297 -3.332538 tp5 | -7.055865 2.279857 -3.09 0.002 -11.5243 -2.587427 tp6 | -7.396141 2.750854 -2.69 0.007 -12.78772 -2.004566 time_trend | -.5605306 .4597118 -1.22 0.223 -1.461549 .3404879 devs_cum | .0989173 .0701613 1.41 0.159 -.0385963 .2364309 sent_cum | .0205145 .0036359 5.64 0.000 .0133882 .0276407 patches_su~d | .053478 .0332304 1.61 0.108 -.0116525 .1186085 norm_indeg~m | 54.67736 17.65584 3.10 0.002 20.07256 89.28217 ------------------------------------------------------------------------------ . . clear . . *The following two models are based on the Postgres data . . use commit_postgres . . stpiece time_trend devs_cum sent_cum patches_submitted norm_indegree_cum, tp(0) cluste > r(id) nohr Invoking stsplit... Creating time pieces... failure _d: commit == 1 analysis time _t: tf id: id Iteration 0: log pseudolikelihood = -158.68819 Iteration 1: log pseudolikelihood = -115.72289 (backed up) Iteration 2: log pseudolikelihood = -94.479168 Iteration 3: log pseudolikelihood = -84.187658 Iteration 4: log pseudolikelihood = -80.279581 Iteration 5: log pseudolikelihood = -80.145847 Iteration 6: log pseudolikelihood = -80.145248 Iteration 7: log pseudolikelihood = -80.145248 Exponential regression -- log relative-hazard form No. of subjects = 3545 Number of obs = 178510 No. of failures = 20 Time at risk = 14877.63753 Wald chi2(6) = 653.56 Log pseudolikelihood = -80.145248 Prob > chi2 = 0.0000 (Std. Err. adjusted for 3545 clusters in id) ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- tp1 | -2.424966 1.564204 -1.55 0.121 -5.490749 .6408161 time_trend | -2.628317 .8242459 -3.19 0.001 -4.243809 -1.012825 devs_cum | 1.028435 .3559904 2.89 0.004 .3307071 1.726164 sent_cum | .0015601 .0006568 2.38 0.018 .0002727 .0028474 patches_su~d | .0459895 .025031 1.84 0.066 -.0030704 .0950494 norm_indeg~m | 65.98327 9.157014 7.21 0.000 48.03585 83.93069 ------------------------------------------------------------------------------ . stpiece time_trend devs_cum sent_cum patches_submitted norm_indegree_cum, tp(0,0.5,1,2 > ,3,4) cluster(id) nohr Invoking stsplit... Creating time pieces... failure _d: commit == 1 analysis time _t: tf id: id Iteration 0: log pseudolikelihood = -158.68819 Iteration 1: log pseudolikelihood = -115.51904 (backed up) Iteration 2: log pseudolikelihood = -90.585747 Iteration 3: log pseudolikelihood = -81.559594 Iteration 4: log pseudolikelihood = -75.924102 Iteration 5: log pseudolikelihood = -75.751924 Iteration 6: log pseudolikelihood = -75.749659 Iteration 7: log pseudolikelihood = -75.749657 Exponential regression -- log relative-hazard form No. of subjects = 3545 Number of obs = 192292 No. of failures = 20 Time at risk = 14877.63753 Wald chi2(11) = 639.82 Log pseudolikelihood = -75.749657 Prob > chi2 = 0.0000 (Std. Err. adjusted for 3545 clusters in id) ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- tp1 | -4.136479 1.950034 -2.12 0.034 -7.958475 -.3144824 tp2 | -4.547613 1.982755 -2.29 0.022 -8.433742 -.6614846 tp3 | -3.420741 1.519339 -2.25 0.024 -6.398591 -.4428917 tp4 | -3.947624 1.472964 -2.68 0.007 -6.834581 -1.060668 tp5 | -5.683403 2.251842 -2.52 0.012 -10.09693 -1.269874 tp6 | -6.681514 2.523801 -2.65 0.008 -11.62807 -1.734954 time_trend | -2.259663 .9057023 -2.49 0.013 -4.034806 -.4845187 devs_cum | .9541991 .4163902 2.29 0.022 .1380894 1.770309 sent_cum | .0021323 .0009993 2.13 0.033 .0001739 .0040908 patches_su~d | .0539797 .0264279 2.04 0.041 .002182 .1057773 norm_indeg~m | 69.36761 11.17426 6.21 0.000 47.46646 91.26877 ------------------------------------------------------------------------------ . . clear . . *The following two models are based on the Python data . . use commit_python . . stpiece time_trend devs_cum sent_cum patches_submitted norm_indegree_cum, tp(0) cluste > r(id) nohr Invoking stsplit... Creating time pieces... failure _d: commit == 1 analysis time _t: tf id: id Iteration 0: log pseudolikelihood = -319.60812 Iteration 1: log pseudolikelihood = -311.14427 Iteration 2: log pseudolikelihood = -286.42189 Iteration 3: log pseudolikelihood = -253.16229 Iteration 4: log pseudolikelihood = -250.84958 Iteration 5: log pseudolikelihood = -250.56169 Iteration 6: log pseudolikelihood = -250.55346 Iteration 7: log pseudolikelihood = -250.55344 Exponential regression -- log relative-hazard form No. of subjects = 1320 Number of obs = 45216 No. of failures = 62 Time at risk = 3768.000883 Wald chi2(6) = 964.93 Log pseudolikelihood = -250.55344 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1320 clusters in id) ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- tp1 | 1.923886 1.540813 1.25 0.212 -1.096052 4.943824 time_trend | -1.377711 .4044375 -3.41 0.001 -2.170394 -.5850282 devs_cum | .10794 .04246 2.54 0.011 .0247199 .19116 sent_cum | .0033191 .0008214 4.04 0.000 .0017092 .004929 patches_su~d | .0943008 .0239296 3.94 0.000 .0473997 .1412018 norm_indeg~m | 8.394936 1.812135 4.63 0.000 4.843216 11.94666 ------------------------------------------------------------------------------ . stpiece time_trend devs_cum sent_cum patches_submitted norm_indegree_cum, tp(0,0.5,1,2 > ,3,4) cluster(id) nohr Invoking stsplit... Creating time pieces... failure _d: commit == 1 analysis time _t: tf id: id Iteration 0: log pseudolikelihood = -319.60812 Iteration 1: log pseudolikelihood = -309.90944 Iteration 2: log pseudolikelihood = -285.83549 Iteration 3: log pseudolikelihood = -251.74784 Iteration 4: log pseudolikelihood = -249.4536 Iteration 5: log pseudolikelihood = -249.17439 Iteration 6: log pseudolikelihood = -249.16598 Iteration 7: log pseudolikelihood = -249.16595 Iteration 8: log pseudolikelihood = -249.16595 Exponential regression -- log relative-hazard form No. of subjects = 1320 Number of obs = 49217 No. of failures = 62 Time at risk = 3768.000883 Wald chi2(11) = 1001.36 Log pseudolikelihood = -249.16595 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1320 clusters in id) ------------------------------------------------------------------------------ | Robust _t | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- tp1 | 1.603189 1.596507 1.00 0.315 -1.525907 4.732285 tp2 | 1.805216 1.715742 1.05 0.293 -1.557577 5.168009 tp3 | 1.596268 1.668 0.96 0.339 -1.672953 4.865488 tp4 | 1.203453 1.662587 0.72 0.469 -2.055158 4.462064 tp5 | 1.123151 1.789364 0.63 0.530 -2.383937 4.63024 tp6 | .8880212 2.134116 0.42 0.677 -3.29477 5.070812 time_trend | -1.355971 .4354636 -3.11 0.002 -2.209464 -.5024785 devs_cum | .1124256 .0456902 2.46 0.014 .0228744 .2019769 sent_cum | .0036775 .0010629 3.46 0.001 .0015944 .0057607 patches_su~d | .0931382 .0241096 3.86 0.000 .0458843 .1403921 norm_indeg~m | 8.539646 1.869068 4.57 0.000 4.876341 12.20295 ------------------------------------------------------------------------------ . . clear . . log close log: D:\Anands Documents\Working Papers\Devanbu\Data\commit\analysis_fse.txt log type: text closed on: 19 Mar 2007, 16:58:07 ----------------------------------------------------------------------------------------