summarise: functions to summarise the pcpt chain results.
In taylors2/PeriodCPT: Within Period Changepoint Detection for Periodic Data

summarise

R Documentation

functions to summarise the pcpt chain results.

Description

Procedures to summarise the MCMC results into a summary table(s) with sufficienct statistics.

Usage

  summarise_chains(object, all = TRUE)
  summarize_chains(object, all = TRUE)

Arguments

`object`	An pcpt class object.
`all`	Logical indicating whether all of the chains should be combined into a single summary table.

Details

The output of PeriodCPT() contains the full MCMC samples per chain within the results slot. The function summarise_chains() creates a frequency table where each row describes a unique sample. The order of the columns are:

m: Number of within period changepoint events.
freq: A tally of the number of times the within period changepoint vector appears in the chain.
tau[j]: The within period changepoint event. An NA cell is used to pad out short changepoint vectors such that the the array can account for changepoint vector that have more events.

The following columns contain the posterior sufficient statistics for the segment parameters for the the corresponding segments depending on the specified sampling distribution.

If all=FALSE, then the table is evaluated for each chain, and the results slot will contain a list of tables in the appropriate order. Otherwise, if all=TRUE then the tables from each chain (whether previously summarised or not) is combined into a single overall table. In this case, the results slot will contain a list with only one item of the overall table.

Note that summarise_chains() changes logical flag of the summarise slot of the pcpt object to indicate that the results slot contains the summarised values rather than the raw chains. Also, the MCMC.last slot is populated with the last sample per chain to permit more MCMC samples to be drawn.

Value

An object of S4 class "pcpt" is returned.

Author(s)

Simon Taylor

Examples

  set.seed(1)
  data = ts( rbinom(120, size = 1, prob = rep(c(0.2, 0.8), each=6)), freq = 12)
  ans  = PeriodCPT.bern(data, niter = 1e3)
  ans  = summarise_chains(ans)
  head(results(ans)[[1]])

taylors2/PeriodCPT documentation built on June 28, 2024, 12:32 p.m.