bcpmeta-package:
Bayesian Multiple Changepoint Detection Using Metadata
Description
Package for a Bayesian multiple changepoint detection method
to detect mean shifts in AR(1) time series.
It accomodates metadta (if available) by
letting metadata times have higher prior probabilities
to be changepoints.
The changepoint configuration with the highest
posterior probability is the optimal model.
Metropolis-Hastings algorithm is utilized for
quick stochastic search of a potentially huge model space.
This method is ideal for annual series,
since it allows a linear trend component, but not yet monthly cycles.
Details
|
Package: |
| bcpmeta |
|
Type: |
| Package |
|
Version: |
| 1.0 |
|
Date: |
| 2014-05-15 |
|
License: |
| GPL(>= 2) |
The most important functions of this package:
bcpmeta.model:
find the optimal changepoint configuration,
bcpmeta.parameters:
given a configuration, estimate model parameters.
References
Li, Y. and Lund, R. (2014) Bayesian Mulitple Changepoint
Detection Using Metadata. (submitted)