suggest_best_k-ConsensusPartitionList-method.RdSuggest the best number of subgroups
# S4 method for ConsensusPartitionList
suggest_best_k(object, jaccard_index_cutoff = select_jaccard_cutoff(ncol(object)))A ConsensusPartitionList-class object.
The cutoff for Jaccard index for comparing to previous k.
It basically gives the best k for each combination of top-value method and partitioning method by calling suggest_best_k,ConsensusPartition-method.
1-PAC score higher than 0.95 is treated as very stable partition (marked by **) and higher than 0.9 is treated as stable partition (marked by *).
A data frame with the best k and other statistics for each combination of methods.
data(golub_cola)
suggest_best_k(golub_cola)
#> best_k 1-PAC mean_silhouette concordance optional_k
#> SD:kmeans 2 1.0000000 0.9748797 0.9791667 **
#> SD:skmeans 2 1.0000000 0.9878910 0.9950000 **
#> CV:kmeans 2 1.0000000 0.9783007 0.9902778 **
#> CV:skmeans 2 1.0000000 0.9754655 0.9894444 **
#> CV:mclust 2 1.0000000 0.9706630 0.9875000 **
#> MAD:kmeans 2 1.0000000 0.9954798 0.9900000 **
#> MAD:skmeans 2 1.0000000 0.9772774 0.9902778 **
#> ATC:skmeans 3 1.0000000 0.9788419 0.9900000 ** 2
#> ATC:mclust 2 1.0000000 0.9684868 0.9819444 **
#> SD:mclust 3 0.9957374 0.9730475 0.9869444 ** 2
#> ATC:pam 3 0.9503951 0.9308092 0.9738889 **
#> ATC:kmeans 2 0.9420544 0.9542907 0.9802778 *
#> MAD:mclust 2 0.9139596 0.9429971 0.9766667 *
#> ATC:hclust 2 0.8867428 0.8665934 0.9472222
#> SD:hclust 2 0.8125549 0.9108634 0.9577778
#> MAD:pam 2 0.8002634 0.9045510 0.9588889
#> SD:pam 2 0.7695347 0.9214171 0.9597222
#> CV:pam 2 0.6931519 0.8694426 0.9394444
#> CV:hclust 2 0.5627744 0.8160426 0.9097222
#> MAD:hclust 2 0.5610184 0.8694379 0.9308333