Importance statistics based on random forests

Computes the difference statistic $$W_j = |Z_j| - |\tilde{Z}_j|$$ where \(Z_j\) and \(\tilde{Z}_j\) are the random forest feature importances of the jth variable and its knockoff, respectively.

stat.random_forest(X, X_k, y, ...)

Arguments

X

n-by-p matrix of original variables.

X_k

n-by-p matrix of knockoff variables.

y

vector of length n, containing the response variables. If a factor, classification is assumed, otherwise regression is assumed.

...

additional arguments specific to ranger (see Details).

Value

A vector of statistics \(W\) of length p.

Details

This function uses the ranger package to compute variable importance measures. The importance of a variable is measured as the total decrease in node impurities from splitting on that variable, averaged over all trees. For regression, the node impurity is measured by residual sum of squares. For classification, it is measured by the Gini index.

For a complete list of the available additional arguments, see ranger::ranger().

See also

Other statistics: stat.SHAP(), stat.forward_selection(), stat.glmnet_coefdiff(), stat.glmnet_lambdadiff(), stat.sqrt_lasso(), stat.stability_selection(), stat.xgboost()

Examples

set.seed(2024)
n=80; p=100; k=10; Ac = 1:k; Ic = (k+1):p
X = generate_X(n=n,p=p)
y <- generate_y(X, p_nn=k, a=3)
Xk = create.shrink_Gaussian(X = X, n_ko = 10)
res1 = knockoff.filter(X, y, Xk, statistic = stat.random_forest,
                       offset = 1, fdr = 0.1)
res1
#> Call:
#> knockoff.filter(X = X, y = y, Xk = Xk, statistic = stat.random_forest, 
#>     fdr = 0.1, offset = 1)
#> 
#> Selected variables:
#> integer(0)
#> 
#> Frequency of selected variables from 0 knockoff copys:
#>   [1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
#>  [38] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
#>  [75] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
perf_eval(res1$shat,Ac,Ic)
#> [1] 0 0