crops {crops} | R Documentation |
Generic implementation of the crops algorithm (ref goes here).
Description
Provides a generic implementation of the crops (changepoints for a range of penalties) algorithm of Haynes et al. (2014) which efficiently searches a range of penalty values in multiple changepoint problems. The crops algorithm finds the optimal segmentations for a different number of segments without incurring as large a computational cost as solving the constrained optimisation problem for a range of values for the number of changepoints. To make the method generic, the user must provide a function that maps a penalty value to the results obtained by a penalised cost changepoint method, and formats these results in a specific way. This interface to the generic method is similar to that as used by the optimx package.
Usage
crops(method, beta_min, beta_max, max_iterations = 20, ...)
Arguments
method |
A function mapping a penalty value to the results obtained by a penalised cost changepoint method. The function must return a list containing the cost and a vector of changepoint locations corresponding to the optimal segmentation as determined by a penalised cost changepoint method. |
beta_min |
A positive numeric value indicating the smallest penalty value to consider. |
beta_max |
A positive numeric value indicating the maximum penalty value to consider. |
max_iterations |
Positive non zero integer. Limits the maximum number of iterations of the crops algorithm to |
... |
Additional parameters to pass to the underlying changepoint method if required. |
Value
An instance of an S4 class of type crops.class
.
References
Haynes K, Eckley IA, Fearnhead P (2017). “Computationally Efficient Changepoint Detection for a Range of Penalties.” Journal of Computational and Graphical Statistics, 26(1), 134-143. doi:10.1080/10618600.2015.1116445.
Nash JC, Varadhan R (2011). “Unifying Optimization Algorithms to Aid Software System Users: optimx for R.” Journal of Statistical Software, 43(9), 1–14. https://www.jstatsoft.org/v43/i09/.
Nash JC (2014). “On Best Practice Optimization Methods in R.” Journal of Statistical Software, 60(2), 1–14. https://www.jstatsoft.org/v60/i02/.
Nash JC (2021). optimx: Expanded Replacement and Extension of the 'optim' Function. R package version 2021-6.12.
Maidstone R, Hocking T, Rigaill G, Fearnhead P (2017). “On optimal multiple changepoint algorithms for large data.” Statistics and Computing, 27. https://link.springer.com/article/10.1007/s11222-016-9636-3.
Rigaill G (2019). fpop: Segmentation using Optimal Partitioning and Function Pruning. R package version 2019.08.26.
Examples
# generate some simple data
set.seed(1)
N <- 100
data.vec <- c(rnorm(N), rnorm(N, 2), rnorm(N))
# example one - calling fpop via crops using global scope
# need the fpop library
library(pacman)
p_load(fpop)
# create a function to wrap a call to fpop for use with crops
fpop.for.crops<-function(beta)
{
# Note - this code is taken from the example in the fpop package
fit <- Fpop(data.vec, beta)
end.vec <- fit$t.est
change.vec <- end.vec[-length(end.vec)]
start.vec <- c(1, change.vec+1)
segs.list <- list()
for(seg.i in seq_along(start.vec))
{
start <- start.vec[seg.i]
end <- end.vec[seg.i]
seg.data <- data.vec[start:end]
seg.mean <- mean(seg.data)
segs.list[[seg.i]] <- data.frame(
start, end,
mean=seg.mean,
seg.cost=sum((seg.data-seg.mean)^2))
}
segs <- do.call(rbind, segs.list)
return(list(sum(segs$seg.cost),segs$end[-length(segs$end)]))
}
# now use this wrapper function with crops
res<-crops(fpop.for.crops,0.5*log(300),2.5*log(300))
# print summary of analysis
summary(res)
# summarise the segmentations
segmentations(res)
# visualise the segmentations
plot(res)
# overlay the data on the segmentations
df <- data.frame("x"=1:300,"y"=data.vec)
plot(res,df)