R Graphical Manual

Browse All

Last data update: 2014.03.03

R: Prediction of warped signals

predict.ptw

R Documentation

Prediction of warped signals

Description

Given a ptw object, predict either the signal at a certain warped time, or the warped time itself.

Usage

## S3 method for class 'ptw'
predict(object, newdata, what = c("response", "time"),
  RTref = NULL, ...)

Arguments

`object`	An object of class "ptw"
`newdata`	Optional vector or matrix of new data points. If `what` equals "response", the new data should be a vector or matrix of intensities. If `what` equals "time", the new data is a vector of time points (a matrix of time points makes no sense...).
`what`	Either "response", in which case the function returns the warped signal, or "time", and then the function returns the warped time axis. That is, the time point in the warped sample corresponding to the given time point in the original sample.
`RTref`	Optional vector of retention times in the reference.
`...`	Further arguments, at the moment not used.

Value

The function returns a matrix (possibly containing only one row) of either warped time points or signals, warped according to the warping function defined in object. When warping signals individually, predict.ptw will check the dimension of newdata: if this is a vector or a matrix of one row, every single warping function will be applied to the one row. If the number of rows equals the number of warping functions, each row will be warped with its corresponding function. If the number of rows does not match the number of warping functions and is not equal to one, an error is given.

Author(s)

Ron Wehrens

References

Eilers, P.H.C. "Parametric Time Warping." Anal. Chem., 2004, 76, 404-411

Bloemberg, T.G. et al. "Improved parametric time warping for proteomics." Chemom. Intell. Lab. Syst., 2010, 104, pp. 65-74

Examples

data(gaschrom)
## Global warping: all samples warped with the same function
ref <- gaschrom[1,]
samp <- gaschrom[14:16,]
gp <- ptw(ref, samp, init.coef = c(0, 1), warp.type = "global")
matplot(t(samp), type = "l", xlim = c(2200, 2400), lty = 1, col = 1:3)
lines(ref, type = "l", col = "gray", lwd = 2)
## plot predicted warped signal directly
matlines(t(predict(gp)), lty = 2, col = 1:3)
## plot original signal at warped time axis
matlines(t(predict(gp, newdata = 2001:2600, what = "time")),
         t(samp[,2001:2600]), col = 1:3, lwd = 3, lty = 2) ## OK

## Individual warping: all samples warped individually
gp <- ptw(ref, samp, init.coef = c(0, 1), warp.type = "indiv")
predict(gp, what = "time", newdata = 2001:2600)
matplot(t(samp), type = "l", xlim = c(2200, 2400), lty = 1, col = 1:3)
lines(ref, type = "l", col = "gray", lwd = 2)
matlines(t(predict(gp, what = "time")),
         t(samp), col = 1:3, lty = 2)

## How would samples 11:13 be warped using the coefficients from samples
## 14:16 (silly but just to make the point)?
samp.pred <- predict(gp, what = "response", newdata = gaschrom[11:13,])

matplot(t(samp), type = "l", lty = 1, col = 1:3, xlim = c(2200, 2400))
lines(ref, type = "l", col = "gray")
matlines(t(samp.pred), type = "l", lty = 2)