Skip to contents

Distance Metrics for Functional Data

Source: R/metric.R
metric.Rd

Functions for computing various distance metrics between functional data. Compute Distance Metric for Functional Data

Usage

metric(fdataobj, fdataref = NULL, method = "lp", ...)

Arguments

fdataobj

An object of class 'fdata'.

fdataref

An object of class 'fdata'. If NULL, computes self-distances.

method

Distance method to use. One of:

  • "lp" - Lp metric (default)

  • "hausdorff" - Hausdorff distance

  • "dtw" - Dynamic Time Warping

  • "pca" - Semi-metric based on PCA scores

  • "deriv" - Semi-metric based on derivatives

  • "basis" - Semi-metric based on basis coefficients

  • "fourier" - Semi-metric based on FFT coefficients

  • "hshift" - Semi-metric with horizontal shift

  • "kl" - Symmetric Kullback-Leibler divergence

...

Additional arguments passed to the specific distance function.

Value

A distance matrix.

Details

Unified interface for computing various distance metrics between functional data objects. This function dispatches to the appropriate specialized distance function based on the method parameter.

This function provides a convenient unified interface for all distance computations in fdars. The additional arguments in ... are passed to the underlying distance function:

  • lp: lp, w

  • hausdorff: (none)

  • dtw: p, w

  • pca: ncomp

  • deriv: nderiv, lp

  • basis: nbasis, basis, nderiv

  • fourier: nfreq

  • hshift: max_shift

  • kl: eps, normalize

Examples

fd <- fdata(matrix(rnorm(200), 20, 10))

# Using different distance methods
D_lp <- metric(fd, method = "lp")
D_hausdorff <- metric(fd, method = "hausdorff")
D_pca <- metric(fd, method = "pca", ncomp = 3)

# Cross-distances
fd2 <- fdata(matrix(rnorm(100), 10, 10))
D_cross <- metric(fd, fd2, method = "lp")