Introduction to Stability Selection
stabilityselectr.RmdThe stabilityselectr package performs stability
selection with a variety of kernels provided by the glmnet
package, and provides simple tools for plotting and extracting selected
features. There is additional functionality designed to facilitate
various forms of permutation clustering analyses.
Useful functions in stabilityselectr
stability_selection()is_stab_sel()get_stable_features()calc_emp_fdr()calc_emp_fdr_breaks()plot_emp_fdr()plot_permuted_data()progeny_cluster()stability_cluster()
Examples
A typical stability selection analysis might be similar to the one below.
Run stability_selection()
set.seed(101)
n_feat <- 20L
n_samples <- 100L
x <- matrix(rnorm(n_feat * n_samples), n_samples, n_feat)
colnames(x) <- paste0("feat", "_", head(letters, n_feat))
y <- sample(1:2, n_samples, replace = TRUE)
stab_sel <- stability_selection(x, y, "binomial", n_iter = 500L)
#> ✓ Using kernel: 'binomial' and 1 core (serial)
#> ✓ Stablity path run time: 1.259s
is_stab_sel(stab_sel)
#> [1] TRUE
stab_sel
#> ══ Stability Selection (Kernel: binomial) ═════════════════════════════
#> • Weakness (alpha) 0.5
#> • Weakness Probability (Pw) 0.2
#> • Number of Iterations 500
#> • Standardized 'Yes'
#> • Imputed Outliers 'No'
#> • Lambda Max 0.2261
#> • Lambda Min Ratio 0.1
#> • Permuted Data 'No'
#> • Random Seed 1234
#> ═══════════════════════════════════════════════════════════════════════
Stable Features at a threshold
get_stable_features(stab_sel, thresh = 0.85)
#> $thresh_0.85
#> # A tibble: 4 × 3
#> feature MaxSelectProb FDRbound
#> <chr> <dbl> <dbl>
#> 1 feat_t 0.917 0.00357
#> 2 feat_d 0.892 0.00714
#> 3 feat_m 0.871 0.0107
#> 4 feat_j 0.866 0.0143
# at multiple thresholds
get_stable_features(stab_sel, thresh = seq(0.7, 0.9, 0.05))
#> $thresh_0.7
#> # A tibble: 20 × 3
#> feature MaxSelectProb FDRbound
#> <chr> <dbl> <dbl>
#> 1 feat_t 0.917 0.00625
#> 2 feat_d 0.892 0.0125
#> 3 feat_m 0.871 0.0188
#> 4 feat_j 0.866 0.025
#> 5 feat_s 0.846 0.0313
#> 6 feat_a 0.824 0.0375
#> 7 feat_g 0.823 0.0438
#> 8 feat_r 0.806 0.05
#> 9 feat_c 0.795 0.0563
#> 10 feat_q 0.795 0.0625
#> 11 feat_l 0.79 0.0688
#> 12 feat_f 0.789 0.075
#> 13 feat_n 0.787 0.0813
#> 14 feat_e 0.761 0.0875
#> 15 feat_h 0.751 0.0938
#> 16 feat_b 0.747 0.1
#> 17 feat_p 0.747 0.106
#> 18 feat_o 0.74 0.113
#> 19 feat_i 0.739 0.119
#> 20 feat_k 0.718 0.125
#>
#> $thresh_0.75
#> # A tibble: 15 × 3
#> feature MaxSelectProb FDRbound
#> <chr> <dbl> <dbl>
#> 1 feat_t 0.917 0.005
#> 2 feat_d 0.892 0.01
#> 3 feat_m 0.871 0.015
#> 4 feat_j 0.866 0.02
#> 5 feat_s 0.846 0.025
#> 6 feat_a 0.824 0.03
#> 7 feat_g 0.823 0.035
#> 8 feat_r 0.806 0.04
#> 9 feat_c 0.795 0.045
#> 10 feat_q 0.795 0.05
#> 11 feat_l 0.79 0.055
#> 12 feat_f 0.789 0.06
#> 13 feat_n 0.787 0.065
#> 14 feat_e 0.761 0.07
#> 15 feat_h 0.751 0.075
#>
#> $thresh_0.8
#> # A tibble: 8 × 3
#> feature MaxSelectProb FDRbound
#> <chr> <dbl> <dbl>
#> 1 feat_t 0.917 0.00417
#> 2 feat_d 0.892 0.00833
#> 3 feat_m 0.871 0.0125
#> 4 feat_j 0.866 0.0167
#> 5 feat_s 0.846 0.0208
#> 6 feat_a 0.824 0.025
#> 7 feat_g 0.823 0.0292
#> 8 feat_r 0.806 0.0333
#>
#> $thresh_0.85
#> # A tibble: 4 × 3
#> feature MaxSelectProb FDRbound
#> <chr> <dbl> <dbl>
#> 1 feat_t 0.917 0.00357
#> 2 feat_d 0.892 0.00714
#> 3 feat_m 0.871 0.0107
#> 4 feat_j 0.866 0.0143
#>
#> $thresh_0.9
#> # A tibble: 1 × 3
#> feature MaxSelectProb FDRbound
#> <chr> <dbl> <dbl>
#> 1 feat_t 0.917 0.00313Progeny and Stability Clustering
See separate vignette on clustering:
vignette("progeny-clustering").