select_order: Calculates different information criterions for a single time...

In marina-khi/multiscale: Multiscale Inference for Nonparametric Time Trend(s)

Calculates different information criterions for a single time series or multiple time series with AR(p) errors based on the long-run variance estimator(s) for a range of tuning parameters and different orders p.

Description

This function fits AR(1), ... AR(9) models for all given time series and calculates different information criterions (FPE, AIC, AICC, SIC, HQ) for each of these fits. The result is the best fit in terms of minimizing the infromation criteria.

Usage

select_order(data, q = NULL, r = 5:15)

Arguments

data	One or a number of time series in a matrix. Column names of the matrix should be reasonable
q	A vector of integers that consisits of different tuning parameters to analyse. If not supplied, q is taken to be [2\log{T}]:([2\sqrt{T}] + 1).
r	A vector of integers that consisits of different tuning parameters r_bar for estimate_lrv. If not supplied, r = 5, \ldots, 15.

Value

A list with a number of elements:

orders	A vector of chosen orders of length equal to the number of time series. For each time series the order is calculated as \max(which.min(FPE), ... which.min(HQ))
...	Matrices with the orders that were selected (among 1, \ldots, 9) for each information criterion. One matrix for each time series.

marina-khi/multiscale documentation built on Jan. 15, 2025, 7:28 a.m.