RECIDICISM FOR RELEASED PRISON INMATES 1 CORRELATIONS OF COVARIATES 22:59 Thursday, November 17, 2005 The CORR Procedure 8 Variables: fin age race wexp mar parole priors educ Pearson Correlation Coefficients, N = 432 Prob > |r| under H0: Rho=0 fin age race wexp fin 1.00000 0.06141 0.06350 0.00468 0.2027 0.1877 0.9228 age 0.06141 1.00000 -0.00271 0.35048 0.2027 0.9552 <.0001 race 0.06350 -0.00271 1.00000 -0.03845 0.1877 0.9552 0.4253 wexp 0.00468 0.35048 -0.03845 1.00000 0.9228 <.0001 0.4253 mar -0.03528 0.17372 -0.05372 0.25234 0.4646 0.0003 0.2653 <.0001 parole -0.01429 -0.07526 0.04004 0.05142 0.7671 0.1183 0.4065 0.2863 Pearson Correlation Coefficients, N = 432 Prob > |r| under H0: Rho=0 mar parole priors educ fin -0.03528 -0.01429 -0.00080 0.05002 0.4646 0.7671 0.9868 0.2996 age 0.17372 -0.07526 -0.10088 0.08008 0.0003 0.1183 0.0361 0.0965 race -0.05372 0.04004 -0.08990 0.09549 0.2653 0.4065 0.0619 0.0473 wexp 0.25234 0.05142 -0.26036 0.18093 <.0001 0.2863 <.0001 0.0002 mar 1.00000 0.09066 -0.02961 0.04004 0.0597 0.5393 0.4065 parole 0.09066 1.00000 -0.12628 0.00961 0.0597 0.0086 0.8421 RECIDICISM FOR RELEASED PRISON INMATES 2 CORRELATIONS OF COVARIATES 22:59 Thursday, November 17, 2005 The CORR Procedure Pearson Correlation Coefficients, N = 432 Prob > |r| under H0: Rho=0 fin age race wexp priors -0.00080 -0.10088 -0.08990 -0.26036 0.9868 0.0361 0.0619 <.0001 educ 0.05002 0.08008 0.09549 0.18093 0.2996 0.0965 0.0473 0.0002 Pearson Correlation Coefficients, N = 432 Prob > |r| under H0: Rho=0 mar parole priors educ priors -0.02961 -0.12628 1.00000 -0.17547 0.5393 0.0086 0.0002 educ 0.04004 0.00961 -0.17547 1.00000 0.4065 0.8421 0.0002 RECIDICISM FOR RELEASED PRISON INMATES 3 A HORIZONTAL BAR CHART 22:59 Thursday, November 17, 2005 week Cum. Cum. Freq Freq Percent Percent | 1 | 1 1 0.23 0.23 2 | 1 2 0.23 0.46 3 | 1 3 0.23 0.69 4 | 1 4 0.23 0.93 5 | 1 5 0.23 1.16 6 | 1 6 0.23 1.39 7 | 1 7 0.23 1.62 8 |* 5 12 1.16 2.78 9 | 2 14 0.46 3.24 10 | 1 15 0.23 3.47 11 | 2 17 0.46 3.94 12 | 2 19 0.46 4.40 13 | 1 20 0.23 4.63 14 | 3 23 0.69 5.32 15 | 2 25 0.46 5.79 16 | 2 27 0.46 6.25 17 | 3 30 0.69 6.94 18 | 3 33 0.69 7.64 19 | 2 35 0.46 8.10 20 |* 5 40 1.16 9.26 21 | 2 42 0.46 9.72 22 | 1 43 0.23 9.95 23 | 1 44 0.23 10.19 24 | 4 48 0.93 11.11 25 | 3 51 0.69 11.81 26 | 3 54 0.69 12.50 27 | 2 56 0.46 12.96 28 | 2 58 0.46 13.43 30 | 2 60 0.46 13.89 31 | 1 61 0.23 14.12 32 | 2 63 0.46 14.58 33 | 2 65 0.46 15.05 34 | 2 67 0.46 15.51 35 | 4 71 0.93 16.44 36 | 3 74 0.69 17.13 37 | 4 78 0.93 18.06 38 | 1 79 0.23 18.29 39 | 2 81 0.46 18.75 40 | 4 85 0.93 19.68 42 | 2 87 0.46 20.14 43 | 4 91 0.93 21.06 44 | 2 93 0.46 21.53 45 | 2 95 0.46 21.99 46 | 4 99 0.93 22.92 47 | 1 100 0.23 23.15 48 | 2 102 0.46 23.61 49 |* 5 107 1.16 24.77 50 | 3 110 0.69 25.46 52 |******************************** 322 432 74.54 100.00 | -----+----+----+----+----+----+-- 50 100 150 200 250 300 Frequency RECIDICISM FOR RELEASED PRISON INMATES 4 COX REGRESSION (DEFAULT BRESLOW TIE CORRECTION) IGNORE TIES AND USE DEFAULT LIKELIHOOD 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling BRESLOW Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1351.367 1316.165 AIC 1351.367 1332.165 SBC 1351.367 1354.054 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 35.2019 8 <.0001 Score 34.9433 8 <.0001 Wald 33.6030 8 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 5 COX REGRESSION (DEFAULT BRESLOW TIE CORRECTION) IGNORE TIES AND USE DEFAULT LIKELIHOOD 22:59 Thursday, November 17, 2005 The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.35928 0.19178 3.5094 0.0610 0.698 age 1 -0.05748 0.02185 6.9202 0.0085 0.944 race 1 0.34566 0.30909 1.2507 0.2634 1.413 wexp 1 -0.11586 0.21300 0.2959 0.5865 0.891 mar 1 -0.42404 0.38202 1.2321 0.2670 0.654 parole 1 -0.09004 0.19567 0.2118 0.6454 0.914 priors 1 0.08438 0.02916 8.3705 0.0038 1.088 educ 1 -0.18554 0.13156 1.9888 0.1585 0.831 RECIDICISM FOR RELEASED PRISON INMATES 6 COX REGRESSION: `EXACT' MODEL TIE CORRECTION SUM LIKELIHOOD OVER ALL POSSIBLE ORDERINGS OF TIE GROUPS 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1192.157 AIC 1227.506 1208.157 SBC 1227.506 1230.046 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 35.3489 8 <.0001 Score 35.0963 8 <.0001 Wald 33.7440 8 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 7 COX REGRESSION: `EXACT' MODEL TIE CORRECTION SUM LIKELIHOOD OVER ALL POSSIBLE ORDERINGS OF TIE GROUPS 22:59 Thursday, November 17, 2005 The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.35963 0.19180 3.5157 0.0608 0.698 age 1 -0.05767 0.02187 6.9563 0.0084 0.944 race 1 0.34556 0.30907 1.2501 0.2635 1.413 wexp 1 -0.11440 0.21312 0.2881 0.5914 0.892 mar 1 -0.42498 0.38209 1.2371 0.2660 0.654 parole 1 -0.08990 0.19568 0.2111 0.6459 0.914 priors 1 0.08471 0.02919 8.4248 0.0037 1.088 educ 1 -0.18577 0.13153 1.9948 0.1578 0.830 RECIDICISM FOR RELEASED PRISON INMATES 8 COX REGRESSION: EFRON TIE CORRECTION 22:59 Thursday, November 17, 2005 EFRON'S LINEAR APPROXIMATION TO `EXACT' MODEL The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EFRON Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1350.761 1315.413 AIC 1350.761 1331.413 SBC 1350.761 1353.302 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 35.3486 8 <.0001 Score 35.0960 8 <.0001 Wald 33.7440 8 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 9 COX REGRESSION: EFRON TIE CORRECTION 22:59 Thursday, November 17, 2005 EFRON'S LINEAR APPROXIMATION TO `EXACT' MODEL The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.35962 0.19180 3.5156 0.0608 0.698 age 1 -0.05767 0.02187 6.9563 0.0084 0.944 race 1 0.34556 0.30907 1.2501 0.2635 1.413 wexp 1 -0.11440 0.21312 0.2882 0.5914 0.892 mar 1 -0.42498 0.38209 1.2371 0.2660 0.654 parole 1 -0.08990 0.19568 0.2111 0.6459 0.914 priors 1 0.08471 0.02918 8.4249 0.0037 1.088 educ 1 -0.18577 0.13153 1.9948 0.1578 0.830 RECIDICISM FOR RELEASED PRISON INMATES 10 COX REGRESSION: `DISCRETE' MODEL TIE CORRECTION CONDITIONAL LOGISTIC REGRESSION AT EACH REARREST TIME 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling DISCRETE Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1192.094 AIC 1227.506 1208.094 SBC 1227.506 1229.983 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 35.4120 8 <.0001 Score 35.1280 8 <.0001 Wald 33.7227 8 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 11 COX REGRESSION: `DISCRETE' MODEL TIE CORRECTION CONDITIONAL LOGISTIC REGRESSION AT EACH REARREST TIME 22:59 Thursday, November 17, 2005 The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.36179 0.19243 3.5349 0.0601 0.696 age 1 -0.05777 0.02191 6.9556 0.0084 0.944 race 1 0.34825 0.31001 1.2619 0.2613 1.417 wexp 1 -0.11680 0.21366 0.2988 0.5846 0.890 mar 1 -0.42614 0.38274 1.2397 0.2655 0.653 parole 1 -0.09064 0.19639 0.2130 0.6444 0.913 priors 1 0.08511 0.02933 8.4230 0.0037 1.089 educ 1 -0.18670 0.13196 2.0019 0.1571 0.830 RECIDICISM FOR RELEASED PRISON INMATES 12 COX REGRESSION: STEPWISE MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Step 1. Variable priors is entered. The model contains the following explanatory variables: priors Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1215.500 AIC 1227.506 1217.500 SBC 1227.506 1220.237 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 12.0053 1 0.0005 Score 14.6407 1 0.0001 Wald 14.3072 1 0.0002 Step 2. Variable age is entered. The model contains the following explanatory variables: RECIDICISM FOR RELEASED PRISON INMATES 13 COX REGRESSION: STEPWISE MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure age priors Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1201.829 AIC 1227.506 1205.829 SBC 1227.506 1211.301 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 25.6769 2 <.0001 Score 25.6277 2 <.0001 Wald 24.5117 2 <.0001 NOTE: No (additional) variables met the 0.05 level for entry into the model. Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio age 1 -0.06933 0.02079 11.1178 0.0009 0.933 priors 1 0.09468 0.02710 12.2007 0.0005 1.099 Summary of Stepwise Selection Variable Number Score Wald Step Entered Removed In Chi-Square Chi-Square Pr > ChiSq 1 priors 1 14.6407 . 0.0001 2 age 2 11.4593 . 0.0007 RECIDICISM FOR RELEASED PRISON INMATES 14 COX REGRESSION: BACKWARDS MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Step 0. The model contains the following variables: fin age race wexp mar parole priors educ Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1192.157 AIC 1227.506 1208.157 SBC 1227.506 1230.046 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 35.3489 8 <.0001 Score 35.0963 8 <.0001 Wald 33.7440 8 <.0001 Step 1. Variable parole is removed. The model contains the following explanatory variables: fin age race wexp mar priors educ RECIDICISM FOR RELEASED PRISON INMATES 15 COX REGRESSION: BACKWARDS MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1192.367 AIC 1227.506 1206.367 SBC 1227.506 1225.520 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 35.1389 7 <.0001 Score 34.9678 7 <.0001 Wald 33.6070 7 <.0001 Step 2. Variable wexp is removed. The model contains the following explanatory variables: fin age race mar priors educ Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1192.674 AIC 1227.506 1204.674 SBC 1227.506 1221.092 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 34.8311 6 <.0001 Score 34.3608 6 <.0001 Wald 33.1264 6 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 16 COX REGRESSION: BACKWARDS MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Step 3. Variable race is removed. The model contains the following explanatory variables: fin age mar priors educ Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1194.066 AIC 1227.506 1204.066 SBC 1227.506 1217.747 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 33.4392 5 <.0001 Score 32.7845 5 <.0001 Wald 31.5474 5 <.0001 Step 4. Variable educ is removed. The model contains the following explanatory variables: fin age mar priors Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1196.098 AIC 1227.506 1204.098 SBC 1227.506 1215.043 RECIDICISM FOR RELEASED PRISON INMATES 17 COX REGRESSION: BACKWARDS MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 31.4077 4 <.0001 Score 31.2458 4 <.0001 Wald 29.9874 4 <.0001 Step 5. Variable mar is removed. The model contains the following explanatory variables: fin age priors Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1198.458 AIC 1227.506 1204.458 SBC 1227.506 1212.667 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 29.0476 3 <.0001 Score 29.0337 3 <.0001 Wald 27.9428 3 <.0001 Step 6. Variable fin is removed. The model contains the following explanatory variables: age priors Convergence Status Convergence criterion (GCONV=1E-8) satisfied. RECIDICISM FOR RELEASED PRISON INMATES 18 COX REGRESSION: BACKWARDS MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1201.829 AIC 1227.506 1205.829 SBC 1227.506 1211.301 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 25.6769 2 <.0001 Score 25.6277 2 <.0001 Wald 24.5117 2 <.0001 NOTE: No (additional) variables met the 0.05 level for removal from the model. Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio age 1 -0.06933 0.02079 11.1178 0.0009 0.933 priors 1 0.09468 0.02710 12.2007 0.0005 1.099 Summary of Backward Elimination Variable Number Wald Step Removed In Chi-Square Pr > ChiSq 1 parole 7 0.2111 0.6459 2 wexp 6 0.3068 0.5796 3 race 5 1.2748 0.2589 4 educ 4 1.9476 0.1628 5 mar 3 2.0453 0.1527 6 fin 2 3.3258 0.0682 RECIDICISM FOR RELEASED PRISON INMATES 19 COX REGRESSION: SCORE MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 RECIDICISM FOR RELEASED PRISON INMATES 20 COX REGRESSION: SCORE MODEL SELECTION (EXACT TIE CORRECTION) 22:59 Thursday, November 17, 2005 The PHREG Procedure Regression Models Selected by Score Criterion Number of Score Variables Chi-Square Variables Included in Model 1 14.6407 priors 1 12.6896 age 1 9.8810 wexp 2 25.6277 age priors 2 19.8938 wexp priors 2 19.0141 fin priors 3 29.0337 fin age priors 3 27.5084 age mar priors 3 27.4447 age priors educ 4 31.2458 fin age mar priors 4 30.6644 fin age wexp priors 4 30.6593 fin age priors educ 5 32.7845 fin age mar priors educ 5 32.5755 fin age race mar priors 5 32.4213 fin age race priors educ RECIDICISM FOR RELEASED PRISON INMATES 21 COX REGRESSION: 22:59 Thursday, November 17, 2005 ADDING A HAZARD FOR EMPLOYMENT DURING THE WEEK OF ARREST The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 432 Number of Observations Used 432 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 432 114 318 73.61 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1227.506 1156.293 AIC 1227.506 1174.293 SBC 1227.506 1198.919 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 71.2123 9 <.0001 Score 66.3165 9 <.0001 Wald 57.8451 9 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 22 COX REGRESSION: 22:59 Thursday, November 17, 2005 ADDING A HAZARD FOR EMPLOYMENT DURING THE WEEK OF ARREST The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.33565 0.19153 3.0713 0.0797 0.715 age 1 -0.04763 0.02152 4.8965 0.0269 0.953 race 1 0.37077 0.31081 1.4230 0.2329 1.449 wexp 1 0.01528 0.21225 0.0052 0.9426 1.015 mar 1 -0.27324 0.38332 0.5081 0.4760 0.761 parole 1 -0.07900 0.19512 0.1639 0.6856 0.924 priors 1 0.07751 0.02952 6.8924 0.0087 1.081 educ 1 -0.20767 0.13338 2.4241 0.1195 0.812 employed 1 -1.33588 0.25062 28.4112 <.0001 0.263 RECIDICISM FOR RELEASED PRISON INMATES 23 ADDING A HAZARD FOR EMPLOYMENT THE WEEK BEFORE THE ARREST 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 431 Number of Observations Used 431 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 431 113 318 73.78 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1215.369 1165.956 AIC 1215.369 1183.956 SBC 1215.369 1208.502 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 49.4132 9 <.0001 Score 48.0202 9 <.0001 Wald 44.8926 9 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 24 ADDING A HAZARD FOR EMPLOYMENT THE WEEK BEFORE THE ARREST 22:59 Thursday, November 17, 2005 The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.33089 0.19223 2.9628 0.0852 0.718 age 1 -0.05056 0.02170 5.4275 0.0198 0.951 race 1 0.35245 0.31024 1.2906 0.2559 1.423 wexp 1 -0.00962 0.21408 0.0020 0.9642 0.990 mar 1 -0.32946 0.38352 0.7379 0.3903 0.719 parole 1 -0.05804 0.19651 0.0872 0.7677 0.944 priors 1 0.08478 0.02935 8.3434 0.0039 1.088 educ 1 -0.19491 0.13308 2.1451 0.1430 0.823 emp_prev 1 -0.79527 0.21803 13.3047 0.0003 0.451 RECIDICISM FOR RELEASED PRISON INMATES 25 EMPLOYMENT HAZARDS FOR THE TWO PREVIOUS WEEKS 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECID Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 430 Number of Observations Used 430 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 430 112 318 73.95 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1203.237 1151.012 AIC 1203.237 1171.012 SBC 1203.237 1198.197 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 52.2245 10 <.0001 Score 49.9122 10 <.0001 Wald 46.5474 10 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 26 EMPLOYMENT HAZARDS FOR THE TWO PREVIOUS WEEKS 22:59 Thursday, November 17, 2005 The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.30986 0.19285 2.5817 0.1081 0.734 age 1 -0.06093 0.02269 7.2125 0.0072 0.941 race 1 0.35426 0.31071 1.2999 0.2542 1.425 wexp 1 0.00566 0.21500 0.0007 0.9790 1.006 mar 1 -0.28819 0.38417 0.5627 0.4532 0.750 parole 1 -0.08200 0.19720 0.1729 0.6775 0.921 priors 1 0.08509 0.02948 8.3333 0.0039 1.089 educ 1 -0.22611 0.13615 2.7582 0.0968 0.798 employ1 1 -0.64882 0.47297 1.8818 0.1701 0.523 employ2 1 -0.15020 0.46790 0.1030 0.7482 0.861 RECIDICISM FOR RELEASED PRISON INMATES 27 HAZARD BASED ON CUMULATIVE AVERAGE NUMBER OF WEEKS EMPLOYED 22:59 Thursday, November 17, 2005 The PHREG Procedure Model Information Data Set WORK.RECIDNEW Dependent Variable week Censoring Variable arrest Censoring Value(s) 0 Ties Handling EXACT Number of Observations Read 431 Number of Observations Used 431 Summary of the Number of Event and Censored Values Percent Total Event Censored Censored 431 113 318 73.78 Convergence Status Convergence criterion (GCONV=1E-8) satisfied. Model Fit Statistics Without With Criterion Covariates Covariates -2 LOG L 1215.369 1175.169 AIC 1215.369 1193.169 SBC 1215.369 1217.716 Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 40.1994 9 <.0001 Score 39.2832 9 <.0001 Wald 37.2806 9 <.0001 RECIDICISM FOR RELEASED PRISON INMATES 28 HAZARD BASED ON CUMULATIVE AVERAGE NUMBER OF WEEKS EMPLOYED 22:59 Thursday, November 17, 2005 The PHREG Procedure Analysis of Maximum Likelihood Estimates Parameter Standard Hazard Variable DF Estimate Error Chi-Square Pr > ChiSq Ratio fin 1 -0.34062 0.19243 3.1332 0.0767 0.711 age 1 -0.05193 0.02199 5.5757 0.0182 0.949 race 1 0.32963 0.30982 1.1319 0.2874 1.390 wexp 1 -0.01339 0.21590 0.0038 0.9505 0.987 mar 1 -0.34763 0.38411 0.8191 0.3654 0.706 parole 1 -0.03692 0.19717 0.0351 0.8515 0.964 priors 1 0.08404 0.02934 8.2060 0.0042 1.088 educ 1 -0.19473 0.13212 2.1723 0.1405 0.823 cumemploy 1 -0.69916 0.30688 5.1908 0.0227 0.497