Faithful: Old Faithful Erruption Interval Data Set
In guenardg/constr.hclust: Space- And Time-Constrained Clustering Package

Faithful

R Documentation

Old Faithful Erruption Interval Data Set

Description

Erruption intervals during a time period of approximately eleven years

Usage

data(Faithful)

Format

A two-column data frame:

Interval: Mean time interval, in hours, between two consecutive erruptions.
Date: Median POSIX time of the averaging window (see below).

Details

Erruption timing data of Old Faithful geyser (Yellowstone National Park, WY, USA) between 2000 and 2011 were downloaded from website http://www.geyserstudy.org (access time: Sun Dec 13 14:43:25 2020). The original data series had 58527 eruption times and showed discontinuities because of interruptions in data recording (caused by equipment malfunction, maintenance, or onboard computer memory overrun caused by harsh weather conditions interferring with satelite data transmission. The series was thus split into 20 segments with continuous observations and the intervals between the eruptions were calculated for each of them.

For exemplary purposes, the segments were decimated to a smaller size by averaging individual time intervals using a 50-eruption wide Hann window and then the decimated data segments were concatenated into a single data series of 1142 estimates of the mean eruption interval.

Source

Guillaume Guenard <guillaume.guenard@gmail.com> and, Pierre Legendre <pierre.legendre@umontreal.ca>; http://www.geyserstudy.org

Examples

data(Faithful)

## Distance matrix (Euclidean) of the mean intervals:

dst <- dist(Faithful$Interval)

## Segmenting the series with respect to eruption time intervals using
## chronological clustering (time-constrained Lance-Williams hierarchical
## agglomerative clustering):

chcl <- constr.hclust(dst, coords = Faithful$Date, chron = TRUE)

### Plotting the results:

## Partition sizes and the colors to display them:
parts <- c(2,3,4,5,6,7)
cols <- list(
  c("red","purple"),
  c("red","blue","purple"),
  c("red","orange","blue","purple"),
  c("red","orange","yellow","blue","purple"),
  c("red","orange","yellow","green","blue","purple"),
  c("red","orange","yellow","green","aquamarine","blue","purple")
)

## Plotting partitions with 2-7 segments:
par(mar=c(5,6.5,2,2))
plot(x=range(Faithful$Date), y=c(0.5,6.5), type="n", xlab="Time",
     ylab="Partitions\n\n", yaxt="n")
for(i in 1L:length(parts)) {
  chcl$coords[,"y"] <- i
  plot(chcl, parts[i], link=TRUE, lwd=25, hybrids="none",
       lwd.pt=0.5, cex=0, pch=21, plot=FALSE, lend=2, col=cols[[i]])
}
axis(2, at=1:length(parts), labels=sprintf("%d groups",parts), las=1)

tmp <- which(!!diff(cutree(chcl,parts[i])))
data.frame(From=Faithful$Date[tmp],
           To=Faithful$Date[tmp+1])

guenardg/constr.hclust documentation built on July 13, 2024, 3:03 p.m.