| segmentation {tilingArray} | R Documentation |
This class represents the result of a segmentation, usually
a call to the function segment.
Objects can be created by calls of the function
segment or by calls of the form new("segmentation", ...).
y:segment.x:nrow(y), or 0. The latter case is equivalent to
x=1:nrow(y).flag:nrow(y), or 0. This can be used to flag certain
probes for special treatment, for example by
plotAlongChrom.breakpoints:breakpoints[[j]] corresponds to a segmentation fit of
j segments, i.e. with j-1 breakpoints. It is a
matrix with (j-1) rows and 1 or 3 columns. It always
contains a column named estimate with the point estimates.
Optionally, it may contain columns lower and upper
with the confidence intervals. The point estimates are the row
indices in y where new segments start, for example:
let z=breakpoints[[j]], then the first segment
is from row 1 to z[1, "estimate"]-1,
the second from row z[1, "estimate"] to
z[2, "estimate"]-1, and so on.negloglik:breakpoints. The
negative log-likelihood of the piecewise
constant models under the data y.hasConfint:breakpoints. TRUE if the confidence interval estimates
are present, i.e. if the matrix breakpoints[[j]] has
columns lower and upper.nrSegments:NA or between 1 and length(breakpoints).
Can be used to select one of the fits in breakpoints for
special treatment, for example by
plotAlongChrom.confint(object, parm, level=0.95,
het.reg=FALSE, het.err=FALSE, ...) computes confidence
intervals for the change point estimates of the
segmentation. Typically, these were obtained from a previous call
to the function segment that created the object.
This is just a wrapper for the function
confint.breakpointsfull
from the strucchange package, which does all the hard
computations.
Parameters: object an object of class segmentation,
parm an integer vector, it determines for which of the segmentation fits
confidence intervals are computed. See also segment.
The other parameters are directly passed on to
confint.breakpointsfull.
logLik(object, penalty="none", ...)
returns the log-likelihoods of fitted models. Valid values for the argument
penalty are none, AIC and BIC.plot(x, y, xlim, xlab="x", ylab="y",
bpcol="black", bplty=1, pch=16, ...)
provides a simple visualization of the result of a
segmentation. Parameters: x an object of class segmentation,
y an integer between 1 and
length(x@breakpoints), selecting which of the fits
contained in x to plot, bpcol and bplty color
and line type of breakpoints. The plot shows the numeric data
along with breakpoints and if available their confidence intervals.Wolfgang Huber huber@ebi.ac.uk
## generate random data with 5 segments: y = unlist(lapply(c(0,3,0.5,1.5,5), function(m) rnorm(10, mean=m))) seg = segment(y, maxseg=10, maxk=15) seg = confint(seg, parm=c(3,4,5)) if(interactive()) plot(seg, 5) show(seg)