Skip to main content

Table 1 The averaged ACVL for several values of K across 100 simulations. Of the 2,000 genes, 50 genes were configured to be informative. The number of patients was 100.

From: Predicting survival outcomes using subsets of significant genes in prognostic marker studies with microarrays

  Correlation matrix
  Exchangeable Lymphoma data*
  ρ = 0.2 ρ = 0.7   
K CC MC CC MC CC MC
1 41.2 41.2 51.6 51.6 48.2 48.2
5 33.3 34.1 46.7 48.2 45.6 46.4
10 29.3 33.2 43.3 46.6 45.2 48.2
20 24.9 49.0 38.4 62.1 45.0 57.4
50 17.3 - 34.6 - 41.7 -
75 21.3 - 40.1 - 42.3 -
100 25.3 - 44.0 - 43.2 -
200 36.0 - 53.3 - 47.0 -
500 50.6 - 63.6 - 53.9 -
1000 56.6 - 70.3 - 58.8 -
  1. *The correlation matrix obtained from the lymphoma data for the top 50 genes in the gene filtering with model (1).
  2. Cox model with the compound covariate (model (3))
  3. Multivariate Cox model (model (6)) for the top K (≤ 20) genes