rfPermute (version 2.5.2)

balancedSampsize: Balanced Sample Size

Description

Create a vector of balanced (equal) sample sizes for use in the sampsize argument of rfPermute or randomForest for a classification model. The values are derived from a percentage of the smallest class sample size.

Usage

balancedSampsize(y, pct = 0.5)

Value

a named vector of sample sizes as long as the number of classes.

Arguments

y

character, numeric, or factor vector containing classes of response variable. Values will be treated as unique for computing class frequencies.

pct

percent of smallest class frequency for sampsize vector.

Author

Eric Archer

Examples

data(mtcars)

# A balanced model with default half of smallest class size
sampsize_0.5 <- balancedSampsize(mtcars$am)
sampsize_0.5

rfPermute(factor(am) ~ ., mtcars, replace = FALSE, sampsize = sampsize_0.5)

# A balanced model with one quarter of smallest class size
sampsize_0.25 <- balancedSampsize(mtcars$am, pct = 0.25)
sampsize_0.25

rfPermute(factor(am) ~ ., mtcars, replace = FALSE, sampsize = sampsize_0.25)