random_forest_gauss: Fits a random forest from a matrix or features X and a vector...

In broadinstitute/cdsr_models: R toolkit for modeling

Fits a random forest from a matrix or features X and a vector y using gausscov for feature selection.

Description

Fits a random forest from a matrix or features X and a vector y using gausscov for feature selection.

Usage

random_forest_gauss(X, y, W = NULL, k = 10, lm = 25, nu = 10, p0 = 0.01)

Arguments

X	n x m numerical matrix of features (missing values will be removed by sample).
y	Length n vector of numerical values (missing values will be removed by column).
W	n row numerical matrix of confounders to be included in each model (after selection).
k	Integer number of cross validation cycles to perform.
lm	The maximum number of linear approximations for gausscov.
nu	The order statistic of Gaussian covariates used for comparison for gausscov.
p0	The P-value cut-off for gausscov.

Value

A list with components:

model_table

A table with the following columns:
feature: the column names of X.
RF.imp.mean: an estimate of the importance of that feature for model accuracy.
RF.imp.sd: the standard deviation of the importance estimate.
RF.imp.stability: the proportion of models that used this feature.
rank: the rank of the feature in terms of importance for the model.
MSE: the mean-squared error of the model.
MSE.se: the standard error of the MSE.
R2: the R^2 of the model.
PearsonScore: the Pearson correlation of predicted and observed responses.

predictions

A vector of the responses predicted by the random forest.

broadinstitute/cdsr_models documentation built on Aug. 9, 2022, 10:36 a.m.