Model 1: GEE-type model fit with exchangable log OR association structure 1 The GENMOD Procedure Response Profile Ordered Total Value y Frequency 1 1 408 2 0 1500 PROC GENMOD is modeling the probability that y='1'. Algorithm converged. GEE Model Information Log Odds Ratio Structure Exchangeable Within-Subject Effect Visit (7 levels) Subject Effect ID (294 levels) Number of Clusters 294 Correlation Matrix Dimension 7 Maximum Cluster Size 7 Minimum Cluster Size 1 Algorithm converged. GEE Fit Criteria QIC 1831.8566 QICu 1822.3096 Analysis Of GEE Parameter Estimates Empirical Standard Error Estimates Standard 95% Confidence Parameter Estimate Error Limits Z Pr > |Z| Intercept -0.5209 0.1215 -0.7591 -0.2827 -4.29 <.0001 Month -0.1712 0.0275 -0.2252 -0.1173 -6.22 <.0001 Month*trt -0.0757 0.0456 -0.1652 0.0138 -1.66 0.0972 Alpha1 3.2294 0.2901 2.6609 3.7979 11.13 <.0001 Model 1: GEE-type model fit with exchangable log OR association structure 2 The GENMOD Procedure Analysis Of GEE Parameter Estimates Model-Based Standard Error Estimates Standard 95% Confidence Parameter Estimate Error Limits Z Pr > |Z| Intercept -0.5209 0.1110 -0.7385 -0.3034 -4.69 <.0001 Month -0.1712 0.0220 -0.2144 -0.1280 -7.77 <.0001 Month*trt -0.0757 0.0361 -0.1465 -0.0049 -2.10 0.0361 Alpha1 3.2294 . . . . . Scale 1.0000 . . . . . NOTE: The scale parameter was held fixed. Score Statistics For Type 3 GEE Analysis Chi- Source DF Square Pr > ChiSq Month 1 40.64 <.0001 Month*trt 1 2.80 0.0941 Model 2: GLMM with random intercepts 3 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 5 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 750.260483 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 743.721075 6.539407 220.2382 -1777.58 2 5 652.996435 90.72464 83.95414 -623.471 3 6 645.972846 7.023588 59.61464 -20.5491 4 7 634.872571 11.10027 27.24724 -46.8505 5 9 630.612833 4.259738 12.3179 -4.30992 6 11 630.12975 0.483084 9.29696 -0.65632 7 13 630.054511 0.075239 3.5136 -0.12475 8 15 630.049231 0.00528 0.796095 -0.00846 9 17 630.048281 0.00095 0.26436 -0.00098 10 19 630.047789 0.000492 0.15187 -0.00028 11 21 630.047776 0.000013 0.013337 -0.00002 12 23 630.047776 4.136E-7 0.00004 -6.87E-7 Model 2: GLMM with random intercepts 4 The NLMIXED Procedure NOTE: GCONV convergence criterion satisfied. Fit Statistics -2 Log Likelihood 1260.1 AIC (smaller is better) 1268.1 AICC (smaller is better) 1268.1 BIC (smaller is better) 1282.8 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.5213 0.2943 293 -5.17 <.0001 0.05 -2.1006 -0.9420 0.000011 b2 -0.3798 0.04224 293 -8.99 <.0001 0.05 -0.4630 -0.2967 0.00004 b3 -0.1387 0.06287 293 -2.21 0.0282 0.05 -0.2624 -0.01491 -0.00003 sigb 3.6899 0.3365 293 10.97 <.0001 0.05 3.0277 4.3522 0.000014 Model 2: GLMM with random intercepts, qpoints=2 5 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 2 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 751.744753 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 745.339634 6.40512 216.7957 -1770.54 2 5 660.529854 84.80978 83.16199 -583.677 3 6 654.398683 6.131172 58.67015 -19.3861 4 7 644.7387 9.659983 32.05753 -43.2184 5 9 641.784427 2.954272 14.05559 -3.45573 6 11 641.470691 0.313737 2.962195 -0.44319 7 13 641.45026 0.02043 0.634042 -0.03227 8 15 641.448714 0.001546 0.97731 -0.00175 9 17 641.44802 0.000694 0.074054 -0.00042 10 19 641.448012 8.92E-6 0.004796 -0.00002 11 21 641.448011 7.451E-8 0.000016 -1.28E-7 NOTE: GCONV convergence criterion satisfied. Model 2: GLMM with random intercepts, qpoints=2 6 The NLMIXED Procedure Fit Statistics -2 Log Likelihood 1282.9 AIC (smaller is better) 1290.9 AICC (smaller is better) 1290.9 BIC (smaller is better) 1305.6 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.4915 0.2722 293 -5.48 <.0001 0.05 -2.0271 -0.9558 -3.68E-6 b2 -0.3606 0.03961 293 -9.11 <.0001 0.05 -0.4386 -0.2827 4.038E-6 b3 -0.1303 0.05892 293 -2.21 0.0277 0.05 -0.2463 -0.01437 -0.00002 sigb 3.2451 0.2487 293 13.05 <.0001 0.05 2.7556 3.7345 1.213E-6 Model 2: GLMM with random intercepts, qpoints=5 7 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 5 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 750.260483 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 743.721075 6.539407 220.2382 -1777.58 2 5 652.996435 90.72464 83.95414 -623.471 3 6 645.972846 7.023588 59.61464 -20.5491 4 7 634.872571 11.10027 27.24724 -46.8505 5 9 630.612833 4.259738 12.3179 -4.30992 6 11 630.12975 0.483084 9.29696 -0.65632 7 13 630.054511 0.075239 3.5136 -0.12475 8 15 630.049231 0.00528 0.796095 -0.00846 9 17 630.048281 0.00095 0.26436 -0.00098 10 19 630.047789 0.000492 0.15187 -0.00028 11 21 630.047776 0.000013 0.013337 -0.00002 12 23 630.047776 4.136E-7 0.00004 -6.87E-7 Model 2: GLMM with random intercepts, qpoints=5 8 The NLMIXED Procedure NOTE: GCONV convergence criterion satisfied. Fit Statistics -2 Log Likelihood 1260.1 AIC (smaller is better) 1268.1 AICC (smaller is better) 1268.1 BIC (smaller is better) 1282.8 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.5213 0.2943 293 -5.17 <.0001 0.05 -2.1006 -0.9420 0.000011 b2 -0.3798 0.04224 293 -8.99 <.0001 0.05 -0.4630 -0.2967 0.00004 b3 -0.1387 0.06287 293 -2.21 0.0282 0.05 -0.2624 -0.01491 -0.00003 sigb 3.6899 0.3365 293 10.97 <.0001 0.05 3.0277 4.3522 0.000014 Model 2: GLMM with random intercepts, qpoints=10 9 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 10 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 750.237502 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 743.69165 6.545852 220.375 -1778.04 2 5 651.938479 91.75317 84.04376 -624.971 3 6 644.283206 7.655274 60.10808 -21.4874 4 7 632.088369 12.19484 27.93273 -50.1662 5 9 626.499802 5.588566 13.9168 -5.204 6 11 625.783263 0.71654 12.17376 -0.92868 7 13 625.630352 0.152911 4.148456 -0.23899 8 15 625.618338 0.012013 0.417315 -0.01882 9 17 625.616695 0.001644 0.107104 -0.00225 10 19 625.616652 0.000042 0.187257 -0.00004 11 20 625.616641 0.000011 0.089498 -0.00002 12 22 625.616617 0.000024 0.038567 -0.00005 13 24 625.616616 1.509E-6 0.00265 -2.41E-6 Model 2: GLMM with random intercepts, qpoints=10 10 The NLMIXED Procedure NOTE: GCONV convergence criterion satisfied. Fit Statistics -2 Log Likelihood 1251.2 AIC (smaller is better) 1259.2 AICC (smaller is better) 1259.3 BIC (smaller is better) 1274.0 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.7191 0.3476 293 -4.95 <.0001 0.05 -2.4033 -1.0350 -0.0002 b2 -0.3907 0.04380 293 -8.92 <.0001 0.05 -0.4769 -0.3044 -0.00018 b3 -0.1432 0.06535 293 -2.19 0.0292 0.05 -0.2718 -0.01457 0.00265 sigb 4.0608 0.4226 293 9.61 <.0001 0.05 3.2291 4.8924 0.000726 Model 2: GLMM with random intercepts, qpoints=20 11 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 20 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 750.237461 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 743.691583 6.545878 220.3755 -1778.04 2 5 651.904366 91.78722 84.06227 -624.976 3 6 644.181625 7.722741 60.14519 -21.5649 4 7 631.857712 12.32391 28.10849 -50.5525 5 9 626.182317 5.675395 14.08522 -5.33096 6 11 625.528621 0.653696 10.51403 -0.87477 7 13 625.422739 0.105882 3.915889 -0.17374 8 15 625.414357 0.008382 0.711921 -0.0141 9 17 625.413448 0.000908 0.182472 -0.00121 10 19 625.413308 0.00014 0.307881 -0.00007 11 21 625.413259 0.00005 0.025329 -0.00006 12 23 625.413257 2.146E-6 0.000101 -3.03E-6 Model 2: GLMM with random intercepts, qpoints=20 12 The NLMIXED Procedure NOTE: GCONV convergence criterion satisfied. Fit Statistics -2 Log Likelihood 1250.8 AIC (smaller is better) 1258.8 AICC (smaller is better) 1258.8 BIC (smaller is better) 1273.6 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.6971 0.3283 293 -5.17 <.0001 0.05 -2.3431 -1.0510 0.000014 b2 -0.3883 0.04325 293 -8.98 <.0001 0.05 -0.4734 -0.3032 0.000101 b3 -0.1424 0.06490 293 -2.19 0.0290 0.05 -0.2701 -0.01464 3.094E-6 sigb 4.0015 0.3745 293 10.68 <.0001 0.05 3.2644 4.7386 0.000027 Model 2: GLMM with random intercepts, qpoints=30 13 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 30 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 750.237461 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 743.691583 6.545878 220.3755 -1778.04 2 5 651.904458 91.78713 84.06223 -624.976 3 6 644.18275 7.721708 60.14443 -21.5643 4 7 631.864017 12.31873 28.10657 -50.5433 5 9 626.210139 5.653878 14.0324 -5.31528 6 11 625.557305 0.652835 10.75016 -0.87447 7 13 625.444915 0.11239 3.941604 -0.18331 8 15 625.436284 0.00863 0.699924 -0.01436 9 17 625.435296 0.000988 0.181987 -0.00131 10 19 625.435156 0.000139 0.314634 -0.00007 11 21 625.435104 0.000052 0.028199 -0.00006 12 23 625.435101 2.771E-6 0.000059 -3.91E-6 Model 2: GLMM with random intercepts, qpoints=30 14 The NLMIXED Procedure NOTE: GCONV convergence criterion satisfied. Fit Statistics -2 Log Likelihood 1250.9 AIC (smaller is better) 1258.9 AICC (smaller is better) 1258.9 BIC (smaller is better) 1273.6 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.6979 0.3304 293 -5.14 <.0001 0.05 -2.3482 -1.0475 0.000017 b2 -0.3885 0.04332 293 -8.97 <.0001 0.05 -0.4738 -0.3033 0.000054 b3 -0.1424 0.06494 293 -2.19 0.0291 0.05 -0.2702 -0.01464 -0.00006 sigb 4.0058 0.3810 293 10.51 <.0001 0.05 3.2560 4.7557 0.000039 Model 2: GLMM with random intercepts, qpoints=50 15 The NLMIXED Procedure Specifications Data Set WORK.ONE Dependent Variable y Distribution for Dependent Variable Binary Random Effects b Distribution for Random Effects Normal Subject Variable ID Optimization Technique Dual Quasi-Newton Integration Method Adaptive Gaussian Quadrature Dimensions Observations Used 1908 Observations Not Used 0 Total Observations 1908 Subjects 294 Max Obs Per Subject 7 Parameters 4 Quadrature Points 50 Parameters b1 b2 b3 sigb NegLogLike -0.5209 -0.1712 -0.0757 1 750.237461 Iteration History Iter Calls NegLogLike Diff MaxGrad Slope 1 3 743.691583 6.545878 220.3755 -1778.04 2 5 651.904457 91.78713 84.06223 -624.976 3 6 644.182721 7.721736 60.14446 -21.5643 4 7 631.863712 12.31901 28.10642 -50.5436 5 9 626.208449 5.655263 14.03569 -5.31637 6 11 625.556098 0.652351 10.70481 -0.87387 7 13 625.445049 0.111049 3.939904 -0.18137 8 15 625.436482 0.008567 0.707442 -0.01429 9 17 625.435509 0.000973 0.184786 -0.00129 10 19 625.435365 0.000144 0.315144 -0.00007 11 21 625.435313 0.000052 0.027606 -0.00006 12 23 625.435311 2.616E-6 0.000073 -3.69E-6 Model 2: GLMM with random intercepts, qpoints=50 16 The NLMIXED Procedure NOTE: GCONV convergence criterion satisfied. Fit Statistics -2 Log Likelihood 1250.9 AIC (smaller is better) 1258.9 AICC (smaller is better) 1258.9 BIC (smaller is better) 1273.6 Parameter Estimates Standard Parameter Estimate Error DF t Value Pr > |t| Alpha Lower Upper Gradient b1 -1.6972 0.3298 293 -5.15 <.0001 0.05 -2.3463 -1.0481 0.000016 b2 -0.3885 0.04330 293 -8.97 <.0001 0.05 -0.4737 -0.3033 0.000073 b3 -0.1424 0.06493 293 -2.19 0.0291 0.05 -0.2702 -0.01464 -0.00004 sigb 4.0044 0.3795 293 10.55 <.0001 0.05 3.2574 4.7513 0.000037