Skip to main content

Table 2 Simulation results on the basis of model (1).

From: Variable selection for large p small n regression models with incomplete data: Mapping QTL with epistases

Model Marker(s) Mean SE BF ≥ 1 BF ≥ 3.2 BF ≥ 10 BF ≥ 100
Without Epistases 2–47.8 0.7453 0.1340 94 (5) 93 (5) 92 (5) 90 (3)
  3–141.5 1.1222 0.231 100 (0) 100 (0) 100 (0) 100 (0)
  FDR (additive) -- -- 0.0067 0 0 0
With Epistases 2–47.8 0.7610 0.1439 94 (6) 93 (6) 93 (6) 93 (6)
  3–141.5 1.1316 0.1402 100 (0) 100 (0) 100 (0) 100 (0)
  (1–24.7, 2–47.8) 1.5607 0.3921 92 (6) 91 (7) 91 (7) 90 (7)
  (2–47.8, 3–141.5) 1.5558 0.3054 97 (3) 96 (4) 96 (4) 96 (4)
  (2–133.8, 3–56.7) -1.6204 0.3875 92 (8) 92 (8) 92 (8) 90 (10)
  FDR (additive) -- -- 0.0408 0.0333 0.0133 0.0067
  FDR (epistatic) -- -- 0.4872 0.3251 0.2283 0.1122
  1. Out of 100 simulated data sets, the total numbers of data sets that correctly identify the true additive and interaction effects (in the brackets, their neighboring ones when the true ones are missed) are counted respectively when thresholding the Bayes factor (BF) at different levels. Also listed are the mean and standard error (SE) of the estimated effect sizes.