Table 2 Simulation results (1000 simulated datasets, each with n patients) showing empirical rejection probabilities of the proposed resampling tests for a classification Random Forest. For all variables the information is given whether or not \(H_0^{(1)}\) and \(H_0^{(2)}\) is true. Thereby, \(H_0^{(1)}\) is considered to be true for variables that are independent to all variables \(X_i\) with non-zero regression coefficient (\(X_3, X_4, X_5\)) and \(H_0^{(2)}\) is considered to be true when the variable’s regression coefficient is zero (\(X_4\), \(X_5\), Z)

From: Exploring the variable importance in random forests under correlations: a general concept applied to donor organ quality in post-transplant survival