hypoRF {hypoRF}

HypoRF; a Random Forest based Two Sample Test

Description

Performs a permutation two sample test based on the out-of-bag-error of random forest

Usage

hypoRF(
  data1,
  data2,
  K = 100,
  statistic = "PerClassOOB",
  normalapprox = F,
  seed = NULL,
  alpha = 0.05,
  ...
)

Arguments

data1

An object of type "data.frame". The first sample.

data2

An object of type "data.frame". The second sample.

K

A numeric value specifying the number of times the created label is permuted. For K = 1 a binomial test is carried out. The Default is K = 100.

statistic

A character value specifying the statistic for permutation testing. Two options available

• PerClassOOB Sum of OOB per class errors.

• OverallOOB OOB-error.

. Default is statistic = "PerClassOOB".

normalapprox

A logical value asking for the use of a normal approximation. Default is normalapprox = FALSE.

seed

A numeric value for reproducibility.

alpha

The level of the test. Default is alpha = 0.05.

...

Arguments to be passed to ranger

Value

A list with elements

• pvalue: The p-value of the test.

• obs: The OOB-statistic in case of K>1 or the out-of-sample error in case of K=1 (binomial test).

• val: The OOB-statistic of the permuted random forests in case of K>1 (otherwise NULL).

• varest: The estimated variance of the permuted random forest OOB-statistic in case of K>1 (otherwise NULL).

• statistic: The used OOB-statistic

• importance_ranking: The variable importance measure, when importance == "impurity".

• cutoff: The quantile of the importance distribution at level alpha.

• call: Call to the function.

See Also

ranger

Examples

# Using the default testing procedure (permutation test)
x1 <- data.frame(x=stats::rt(100, df=1.5))
x2 <- data.frame(x=stats::rnorm(100))
hypoRF(x1, x2, K=2)
# Using the exact binomial test
hypoRF(x1, x2, K=1)