aic: AIC and BIC guided aggregation of segments

In rseg: recursive segmentation (rseg)

Description

Functions to aggregate the last segments in a rseg object in order to optimize the AIC or BIC.

Usage

aic(x, criterion = c("AIC", "BIC")[1], family = c("gaussian",
  "binomial", "censored")[1])

bic(x, family = c("gaussian", "binomial", "censored")[1])

Arguments

x	an object of class rseg fit by cseg, eseg or rseg.
criterion	type of information criterion used. Accepted values are "AIC" (= default) and "BIC".
family	a description of the error distribution/scale of the outcome. Accepted values are "gaussian", "binomial" and "censored".

Details

Depending on the value of the family argument, the AIC and BIC are extracted from a generalized linear model with identity link function (= "gaussian"), logit link function (= "binomial") or a Cox proportional hazards regression model (= "censored"). The segmentation provided by x is used as factor variable in these models.

Value

An object of class rseg.

Examples

airq <- subset(airquality, !is.na(Ozone))
irisseg <- cseg(Ozone ~ ., minsplit = 10L, minbucket = 3L, data = airq)
irisseg

## The AIC does not suggest aggregation of segments
aic(irisseg, family = 'gaussian')

## The stricter BIC suggests aggregation of segments
aic(irisseg, criterion = 'BIC', family = 'gaussian')
## This is equivalent to
bic(irisseg, family = 'gaussian')

rseg documentation built on May 2, 2019, 6:38 p.m.