targetIntervals: Compute target intervals

In penaltyLearning: Penalty Learning

Compute target intervals

Description

Compute target intervals of log(penalty) values that result in predicted changepoint models with minimum incorrect labels. Use this function after labelError, and before IntervalRegression*.

Usage

targetIntervals(models, 
    problem.vars)

Arguments

models	data.table with columns errors, min.log.lambda, max.log.lambda, typically labelError()$model.errors.
problem.vars	character: column names used to identify data set / segmentation problem.

Value

data.table with columns problem.vars, one row for each segmentation problem. The "min.log.lambda", and "max.log.lambda" columns give the largest interval of log(penalty) values which results in the minimum incorrect labels for that problem. This can be used to create the target.mat parameter of the IntervalRegression* functions.

Author(s)

Toby Dylan Hocking <[e-mail]> [aut, cre]

Examples

data.table::setDTthreads(1)

library(penaltyLearning)
data(neuroblastomaProcessed, envir=environment())
targets.dt <- targetIntervals(
  neuroblastomaProcessed$errors,
  problem.vars=c("profile.id", "chromosome"))

penaltyLearning documentation built on Oct. 2, 2024, 9:09 a.m.