Skip to contents

Bayesian Pairwise Curve Alignment

Source: R/alignment.R
bayesian.align.pair.Rd

Align two curves using a Bayesian framework with MCMC sampling of the warping function posterior. This provides uncertainty quantification for the alignment via posterior samples of the warping function.

Usage

bayesian.align.pair(
  f1,
  f2,
  argvals,
  n.samples = 2000,
  burn.in = 500,
  step.size = 0.1,
  proposal.variance = 1,
  seed = 42
)

Arguments

f1

Numeric vector of the first curve's values.

f2

Numeric vector of the second curve's values.

argvals

Numeric vector of the evaluation grid (common to both curves).

n.samples

Number of MCMC samples to draw (default 2000).

burn.in

Number of initial samples to discard as burn-in (default 500).

step.size

Step size for the MCMC proposal (default 0.1).

proposal.variance

Variance of the MCMC proposal distribution (default 1.0).

seed

Random seed for reproducibility (default 42).

Value

A list with components:

gamma.mean

numeric vector of the posterior mean warping function

gamma.samples

matrix of posterior warping function samples (samples x grid)

f2.aligned

numeric vector of the aligned second curve

acceptance.rate

the MCMC acceptance rate

References

Cheng, W., Dryden, I.L., and Huang, X. (2016). Bayesian registration of functions and curves. Bayesian Analysis, 11(2):447–475.

See also

elastic.align

Examples

# \donttest{
t <- seq(0, 1, length.out = 100)
f1 <- sin(2 * pi * t)
f2 <- sin(2 * pi * (t^1.3))
res <- bayesian.align.pair(f1, f2, t, n.samples = 500, burn.in = 100)
# }