Variable selection with Minimal Depth (MD) — var.select.md • RFSurrogates

This function executes MD applying ranger for random forests generation and is a reimplementation of var.select from randomForestSRC package.

Usage

var.select.md(
  x = NULL,
  y = NULL,
  num.trees = 500,
  type = "regression",
  mtry = NULL,
  min.node.size = 1,
  num.threads = NULL,
  status = NULL,
  save.ranger = FALSE,
  create.forest = is.null(forest),
  forest = NULL,
  save.memory = FALSE,
  case.weights = NULL
)

Arguments

x: data.frame of predictor variables with variables in columns and samples in rows. (Note: missing values are not allowed)
y: vector with values of phenotype variable (Note: will be converted to factor if classification mode is used). For survival forests this is the time variable.
num.trees: Number of trees. Default is 500.
type: Mode of prediction ("regression","classification" or "survival"). Default is regression.
mtry: Number of variables to possibly split at in each node. Default is no. of variables^(3/4) as recommended by Ishwaran.
min.node.size: Minimal node size. Default is 1.
num.threads: number of threads used for parallel execution. Default is number of CPUs available.
status: status variable, only applicable to survival data. Use 1 for event and 0 for censoring.
save.ranger: Set TRUE if ranger object should be saved. Default is that ranger object is not saved (FALSE).
create.forest: Default: TRUE if forest is NULL, FALSE otherwise. Whether to create or use an existing forest.
forest: the random forest that should be analyzed.
save.memory: Use memory saving (but slower) splitting mode. No effect for survival and GWAS data. Warning: This option slows down the tree growing, use only if you encounter memory problems. (This parameter is transfered to ranger)
case.weights: Weights for sampling of training observations. Observations with larger weights will be selected with higher probability in the bootstrap (or subsampled) samples for the trees.

Value

List with the following components:

info: list with results from mindep() function:
- depth: mean minimal depth for each variable.
- selected: variables has been selected (1) or not (0).
- threshold: the threshold that is used for the selection. (deviates slightly from the original implementation)
var: vector of selected variables.
forest: a list containing:
- trees: list of trees that was created by getTreeranger(), addLayer(), and addSurrogates() functions and that was used for surrogate minimal depth variable importance.
- allvariables: all variable names of the predictor variables that are present in x.
ranger: ranger object

References

Ishwaran, H. et al. (2011) Random survival forests for high-dimensional data. Stat Anal Data Min, 4, 115–132. https://onlinelibrary.wiley.com/doi/abs/10.1002/sam.10103
Ishwaran, H. et al. (2010) High-Dimensional Variable Selection for Survival Data. J. Am. Stat. Assoc., 105, 205–217. http://www.ccs.miami.edu/~hishwaran/papers/IKGML.JASA.2010.pdf

Examples


# \donttest{
data("SMD_example_data")
set.seed(42)
res <- var.select.md(
  x = SMD_example_data[, 2:ncol(SMD_example_data)],
  y = SMD_example_data[, 1], num.trees = 10, num.threads = 1
)
res$var
#>  [1] "X2"      "X3"      "X4"      "X5"      "X6"      "X8"      "cp1_4"  
#>  [8] "cp2_1"   "cp2_3"   "cp2_5"   "cp2_10"  "cp3_2"   "cp3_3"   "cp3_4"  
#> [15] "cp3_5"   "cp3_6"   "cp7_7"   "cp7_10"  "cp8_1"   "cp8_7"   "cp8_10" 
#> [22] "cp9_4"   "cp9_7"   "cp9_9"   "cp9_10"  "cgn_3"   "cgn_4"   "cgn_6"  
#> [29] "cgn_15"  "cgn_16"  "cgn_17"  "cgn_20"  "cgn_24"  "cgn_35"  "cgn_43" 
#> [36] "cgn_44"  "cgn_47"  "cgn_48"  "cgn_49"  "cgn_51"  "cgn_53"  "cgn_55" 
#> [43] "cgn_58"  "cgn_59"  "cgn_63"  "cgn_68"  "cgn_69"  "cgn_72"  "cgn_78" 
#> [50] "cgn_79"  "cgn_81"  "cgn_82"  "cgn_91"  "cgn_93"  "cgn_94"  "cgn_95" 
#> [57] "cgn_99"  "cgn_101" "cgn_107" "cgn_108" "cgn_110" "cgn_112" "cgn_113"
#> [64] "cgn_115" "cgn_116" "cgn_117" "cgn_120" "cgn_122" "cgn_125" "cgn_126"
#> [71] "cgn_128" "cgn_131"
# }