frmselect: fractional model selection function based on quasi-likelihood

In Lydia2kkx/frmselection: Model Selection Tools for Fractional Response Models

Description

frmselect is used to select a formula-based model by information criterions of fractional regression models.

Usage

frmselect(x, y, criterion="AIC",linkfrac="logit", method="forward", plotit=FALSE)

Arguments

x	a numeric matrix, with column names, containing the values of the covariates.
y	a numeric vector containing the values of the response variable. It should be between 0 and 1.
criterion	model selection critetion. Available options: AIC, BIC, HQ.The default value is AIC.
linkfrac	link function, Available options: logit, probit, loglog, cloglog, cauchit.The default value is logit.
method	the mode of stepwise search and allsubsets. Available options: forward, backward, both, allsubsets.The default value is forward.
plotit	Default value is FALSE. If TRUE, it would plot the number of variables vs minimal criteria.

Value

A list contains information criterion, link function, model selection method, minimal criteria of corresponding number of variables, minimal information criterion, the order of variables which are chosen in the model, the names of corresponding variables and the estimated coefficients of them.

Examples

set.seed(123)
library(bamlss)
d <- GAMart()
x <- as.matrix(data.frame(d$x1, d$x2, d$x3))
y <- d$bnum
frmselect(x, y)

Lydia2kkx/frmselection documentation built on May 27, 2019, 2:05 p.m.