The leaps() function is used to perform the all-subsets regression and defined under the leaps package.
In R, the leaps() function is part of the “leaps” package and is used for model selection using various types of leaps-and-bounds algorithms. It is commonly employed in the context of linear regression to explore all possible subsets of predictors and identify the best subset based on a specified criterion, such as the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), or adjusted R-squared.
Here is a brief explanation of the key parameters and purpose of the leaps() function:
leaps(x, y = NULL, wt = NULL, method = c(“Cp”, “adjr2”, “bic”, “r2”),
nbest = 10, nvmax = NULL, force.in = NULL, force.out = NULL,
int = TRUE, strict = c(“none”, “both”, “in”, “out”),
trace = TRUE, keep.in = FALSE, keep.out = FALSE, …)
x: A matrix or data frame of predictors.
y: A response variable (optional for some methods).
wt: Optional weights for observations.
method: The information criterion to be used for model selection. Options include “Cp” (Mallows’ Cp), “adjr2” (adjusted R-squared), “bic” (Bayesian Information Criterion), and “r2” (R-squared).
nbest: The number of best models to be retained.
nvmax: The maximum number of variables in the final model.
force.in: A vector of variables to be included in all models.
force.out: A vector of variables to be excluded from all models.
int: Logical. Should intercept be included in models?
strict: Specifies the type of constraints to be applied on the models.
trace: Logical. Should diagnostic information be printed during the computation?
keep.in and keep.out: Logical. Should the models with forced-in or forced-out variables be retained?
…: Additional arguments passed to or from methods.
The function returns an object of class “leaps” containing information about the best models, including their coefficients, RSS (Residual Sum of Squares), and the chosen information criterion.
Here’s a simple example:
library(leaps)
# Generate example data
set.seed(123)
x <- matrix(rnorm(100), ncol = 5)
y <- x[, 1] + 2 * x[, 2] + rnorm(100)
# Run leaps with Cp criterion
leaps_result <- leaps(x, y, method = “Cp”)
# Display results
summary(leaps_result)
This example uses the Cp criterion to select the best subset of predictors based on the Mallows’ Cp statistic. Adjust the function parameters according to your specific requirements and dataset.