Skip to main content

Table 2 Number of classes obtained for various clustering methods applied to simulated data

From: Model-based clustering of DNA methylation array data: a recursive-partitioning algorithm for high-dimensional data arising as a mixture of beta distributions

  Case 1 (5 true classes)   Case 2 (4 true classes)  
Method J Median Mean SD J Median Mean SD
DynTree 25 3 2.5 0.50 25 2 2.0 0.00
  50 3 2.5 0.50 50 2 2.0 0.00
  500 3 2.7 0.58 500 2 2.0 0.00
  1000 3 2.8 0.59 1000 2 2.0 0.00
HOPACH (best) 25 40 38.0 12.10 5 17 18.9 9.10
  50 35 35.4 11.38 10 14 15.0 8.27
  500 23 23.0 9.52 25 25 24.7 9.80
  1000 23 23.1 9.47 50 25 25.3 7.34
HOPACH (greedy) 25 8 13.4 14.41 5 5 7.1 6.35
  50 6 11.9 12.66 10 5 7.1 7.11
  500 5 6.6 5.19 25 7.5 10.8 8.52
  1000 4 6.2 4.41 50 8 10.1 7.85
RPMM 25 8 7.7 2.00 5 2 2.0 0.10
  50 5 5.6 1.32 10 2 2.4 2.28
  500 5 5.0 0.22 25 4 4.0 0.20
  1000 5 5.0 0.00 50 4 4.1 0.58
  1. DynTree = Hierarchical clustering with classes determined by dynamic tree cutting
  2. HOPACH(best) = HOPACH with 'best' number of classes
  3. HOPACH(greedy) = HOPACH with 'greedy' number of classes
  4. RPMM = Recursively partitioned mixture model employing BIC
  5. J = Number of loci considered in analysis