Skip to main content

Advertisement

Table 2 Simulation results for six treatment effect (TE)/visit process (VP) scenarios.

From: A coarsened multinomial regression model for perinatal mother to child transmission of HIV

  Bias MSE CP Power
Scenario A1 A2/A3 A1 A2/A3 A1 A2/A3 A1 A2/A3
TE1/VP1 (β 11 = -0.55, β 21 = β 2|1 - ,1 = -0.54)
L-CUM 0.013 0.045 0.069 0.030 0.948 0.943 0.518 0.845
CM-CUM 0.027 0.048 0.068 0.027 0.948 0.939 0.499 0.882
L-COND -0.015 0.107 0.077 0.077 0.959 0.914 0.539 0.439
CM-COND -0.007 0.106 0.072 0.072 0.956 0.901 0.541 0.461
TE2/VP1 (β 11 = -0.55, β 21 = β 2|1 - ,1 = 0)
L-CUM 0.173 -0.020 0.096 0.022 0.879 0.952 0.309  
CM-CUM 0.168 -0.009 0.094 0.020 0.873 0.952 0.339  
L-COND 0.098 0.004 0.083 0.050 0.938 0.956 0.398  
CM-COND 0.101 0.006 0.079 0.046 0.941 0.952 0.405  
TE3/VP1 (β 11 = -0.02, β 21 = β 2|1 - ,1 = -0.38)
L-CUM -0.078 0.055 0.058 0.029 0.952 0.931 0.057 0.517
CM-CUM -0.064 0.049 0.054 0.025 0.954 0.933 0.056 0.605
L-COND -0.050 0.064 0.062 0.059 0.946 0.941 0.062 0.263
CM-COND -0.042 0.063 0.058 0.055 0.944 0.944 0.057 0.264
TE4/VP1 (β 11 = 0.27, β 21 = β 2|1 - ,1 = 0.27)
L-CUM -0.123 0.059 0.065 0.028 0.921 0.936 0.096 0.271
CM-CUM -0.105 0.049 0.057 0.023 0.931 0.942 0.111 0.307
L-COND -0.043 0.034 0.058 0.060 0.953 0.937 0.160 0.173
CM-COND -0.038 0.033 0.055 0.056 0.954 0.936 0.186 0.187
TE4/VP2
L-CUM -0.110 0.116 0.064 0.045 0.927 0.898 0.103 0.138
CM-CUM -0.096 0.043 0.059 0.032 0.927 0.940 0.128 0.261
L-COND -0.073 0.041 0.061 0.084 0.941 0.949 0.138 0.123
CM-COND -0.066 0.042 0.054 0.078 0.946 0.945 0.146 0.134
TE4/VP3
L-CUM -0.124 0.192 0.113 0.105 0.923 0.890 0.068 0.064
CM-CUM -0.088 0.017 0.097 0.058 0.939 0.954 0.096 0.193
L-COND -0.054 -0.009 0.099 0.292 0.944 0.967 0.095 0.077
CM-COND -0.040 0.002 0.087 0.211 0.946 0.964 0.102 0.086
  1. TE scenarios are defined by β 11, β 21, and β 2|1 - ,1, which represent the true log odds ratios for in utero transmission (A1), perinatal transmission (A2), and intrapartum transmission (A3), respectively, comparing treatment versus control (see Appendix A). VP scenarios are defined such that the amount of missing data increases from VP1 to VP3 (see Appendix B). L-CUM = logistic model, cumulative probabilities; CM-CUM = coarsened multinomial model, cumulative probabilities; L-COND = logistic model, conditional probabilities; CM-COND = coarsened multinomial model, conditional probabilities. Results for estimators of treatment effect on perinatal transmission and intrapartum transmission are combined under the heading A2/A3.