fs.reg: Variable selection in generalised linear regression models...

In Rfast: A Collection of Efficient and Extremely Fast R Functions

Variable selection in generalised linear regression models with forward selection

Description

Variable selection in generalised linear regression models with forward selection

Usage

fs.reg(y, ds, sig = 0.05, tol = 2, type = "logistic") 

Arguments

y	The dependent variable. This can either be a binary numeric (0, 1) or a vector with integers (numeric or integer class), count data. The first case is for the binary logistic regression and the second for the Poisson regression.
ds	The dataset; provide a matrix where columns denote the variables and the rows the observations. The variables must be continuous, no categorical variables are accepted.
sig	Significance level for assessing the p-values significance. Default value is 0.05.
tol	The difference bewtween two successive values of the stopping rule. By default this is is set to 2. If for example, the BIC difference between two succesive models is less than 2, the process stops and the last variable, even though significant does not enter the model.
type	If you have a binary dependent variable, put "logistic" or "quasibinomial". If you have percentages, values between 0 and 1, including 0 and or 1, use "quasibinomial" as well. If you have count data put "poisson".

Details

The classical forward regression is implemented. The difference is that we have an extra step of check. Even if a variable is significant, the BIC of the model (with that variable) is calculated. If the decrease from the previous BIC (of the model without this variable) is less thatn a prespecified by the user value (default is 2) the variable wil enter. This way, we guard somehow against over-fitting.

Value

A matrix with for columns, the selected variables, the logarithm of their p-value, their test statistic and the BIC of the model with these variables included. If no variable is selected, the matrix is empty.

Author(s)

Marios Dimitriadis

Documentation: Marios Dimitriadis <[e-mail]>.

See Also

cor.fsreg, logistic_only, poisson_only, glm_logistic, glm_poisson

Examples

set.seed(123)

#simulate a dataset with continuous data
x <- matrnorm(100, 50)
y <- rpois(100, 10)
a <- fs.reg(y, x, sig = 0.05, tol = 2, type = "poisson") 
x <- NULL

Rfast documentation built on Nov. 9, 2023, 5:06 p.m.