fselect: F-value stepwise Model selection

In hans0803/APLM:

F-value stepwise Model selection

Description

do linear model from data and y, and use F-value selection to choose model return. can select forward or backward run this function, alpha value decide p-value select accuracy. and if model have Multicollinearity problem, function will warning it.

Usage

fselect(y, data, alpha_in, alpha_out, mode)

Arguments

y	y of linear model, a strain variable
data	all x of linear model, independent variable
alpha_in	new variable p-value should smaller than alpha in, it can be select into model
alpha_out	if p-value bigger than alpha out, the variable will be dropout to model
mode	can choose forward or backward stepwise selection

Examples

y <- iris[,1]
x <- iris[,2:4]
fit <- fselect(y=y, data=x, alpha_in=0.01, alpha_out=0.05, mode="backward")
summary(fit)

y <- MASS::Boston[,14]
x <- MASS::Boston[,-14]
fit <- fselect(y=y, data=x, alpha_in=0.01, alpha_out=0.05, mode="forward")
summary(fit)

data <- as.data.frame(matrix(rnorm(1000), ncol=10))
data[,2] <- data[,1]*5+rnorm(100)
# beta <- c(beta_0, beta_1, ..., beta_n)
beta <- c(4,2,3,0,1,7,0,5)
# use function set y
y <- yget(data, beta)
mult_fit <- fselect(y, data, alpha_in=0.001, alpha_out=0.005)
summary(mult_fit)

hans0803/APLM documentation built on June 17, 2024, 5:22 a.m.