Skip to main content

Table 3 The proportions of simulations that having quasi-complete separation issues among their 1000 bootstraps

From: How to analyze work productivity loss due to health problems in randomized controlled trials? A simulation study

Distribution of productivity loss outcome Number of observations in each arm Number of quasi-complete separation detected in 1000 bootstraps
0 1–10 11–100 101–400 401–700 701–950 >  950
80:15:5/60:30:10 100 38.03 41.97 16.64 3.25 0.01 0.01 0.09
80:15:5/60:30:10 200 94.75 5.00 0.24 0.01 0 0 0
60:35:5/40:50:10 100 61.16 32.83 5.63 0.38 0 0 0
60:35:5/40:50:10 200 99.17 0.81 0.02 0 0 0 0
50:40:10/30:55:15 50 39.45 45.05 14.30 1.11 0.08 0.01 0
50:40:10/30:55:15 100 97.46 2.46 0.08 0 0 0 0
50:40:10/30:55:15 200 100 0 0 0 0 0 0
  1. Among 10000 simulations for each row: 5000 for equal scale and 5000 for unequal scale for the truncated negative binomial distributions of productivity loss outcomes in the two arms