Skip to main content

Advertisement

Table 3 Simulation results under H 0 and H 1 of the tests used for comparative purposes

From: Multivariate permutation test to compare survival curves for matched data

Scenario* Strata effect (w) N. of strata (k) Under H0 Under H1
    (1) (2) (3) (1) (2) (3)
PH 0 30 0.042 0.043 0.049 0.660 0. 734 0. 770
  0 50 0.042 0.043 0.051 0.867 0. 923 0. 938
  0 100 0.054 0.050 0.060 0.995 0. 999 1. 000
  0.22 30 0.044 0.052 0.053 0.635 0. 731 0. 726
  0.22 50 0.057 0.051 0.058 0.827 0. 898 0. 903
  0.22 100 0.056 0.054 0.053 0.986 0. 997 0. 996
ED 0 30 0.053 0.061 0.064 0.207 0.189 0.216
  0 50 0.061 0.067 0.071 0.317 0.262 0.310
  0 100 0.049 0.052 0.061 0.620 0.531 0.571
  0.22 30 0.046 0.051 0.056 0.204 0.169 0.166
  0.22 50 0.037 0.044 0.045 0.292 0.237 0.253
  0.22 100 0.051 0.048 0.053 0.596 0.501 0.510
LD 0 30 0.042 0.046 0.054 0.384 0.473 0.520
  0 50 0.053 0.059 0.060 0.600 0.731 0.775
  0 100 0.052 0.052 0.118 0.863 0.965 0.971
  0.22 30 0.049 0.065 0.070 0.338 0.454 0.470
  0.22 50 0.046 0.051 0.051 0.553 0.707 0.705
  0.22 100 0.051 0.056 0.051 0.816 0.949 0.947
CH 0 30 0.062 0.070 0.076 0.076 0.078 0.082
  0 50 0.045 0.045 0.046 0.087 0.119 0.119
  0 100 0.041 0.045 0.048 0.095 0.174 0.662
  0.22 30 0.046 0.043 0.049 0.079 0.091 0.083
  0.22 50 0.040 0.042 0.039 0.101 0.142 0.138
  0.22 100 0.063 0.056 0.064 0.116 0.233 0.233
  1. * PH=Proportional Hazards, ED=Early Difference, LD=Late Difference, CH=Crossing Hazards.
  2. The outcomes of the stratified log-rank test (1), the modified log-rank by Schoenfeld and Tsiatis for highly stratified data (2) and the Cox model with sandwich variance that accounts for stratified data with possibly correlated failure times (3) are reported. In the body of the table, values in bold identify the strongest results.