fixNA.Rd

\name{fixNA}
\alias{fixNA}
%- Also NEED an '\alias' for EACH other topic documented here.
\title{
Impute missing values into a column of amplification data
}
\description{
The function \code{\link[chipPCR]{fixNA}} imputes missing values in a single 
column of data. The imputation is based on a linear approximation by default. 
However, the data can also be estimated from an approximation by splines.
}
\usage{
fixNA(x, y, spline = TRUE, verbose = FALSE)
}
%- maybe also 'usage' for other objects documented here.
\arguments{
  \item{x}{
    \code{x} is a vector containing the time or cycle values.
}
  \item{y}{
    \code{y} is a vector containing the fluorescence values.
}
  \item{spline}{
	\code{spline} is a logical operator which defines of missing values 
	should be imputed by spline approximation as per 
	\code{\link[stats]{spline}}. The default is linear a approximation as 
	per \code{\link[stats]{approx}}.
}
  \item{verbose}{
    \code{verbose} defines if the number of missing values should be reported.
}
}
\details{
Amplification data of experimental systems may contain missing values (NA). 
The NAs may be caused by detector problems, acquisition error or other 
assorted problems. There are different ways to handle missing values. One 
approach is to ignore NAs which is generally acceptable. However, in case of 
further calculation it is often necessary to handle cases of missing values in 
a way that the next calculation steps can be performed. Missing values can be 
eliminated by a imputation. Imputation encompasses various approaches. This 
includes to calculate a location parameter (e.g., mean, median) or other 
significant values (e.g., minimum, maximum, modus) of a data column. However, 
in non-linear processes such as amplification processes its is better to 
estimate the missing values from a trend. The function 
\code{\link[chipPCR]{fixNA}} was empirically tested and relies on a linear 
trend estimation based on the \code{\link[stats]{approx}} function. This 
approach is useful but may be problematic on the phases other then background 
or plateau phases of an amplification reaction. The parameter 
\code{\link[stats]{spline}} on \code{\link[chipPCR]{fixNA}} enables a trend 
estimation on splines and may be more appropriate in most scenarios. Other 
smooting functions such as the Savitzky-Golay smoothing filter have the 
intrinsic capability to remove missing values [Savitzk and Golay 1964, 
Eilers 2003].
}

\references{
  Eilers, P. H. C. Anal. Chem. 2003, 75, 3631--3636.
  Savitzky, A.; Golay, M. J. E. Anal. Chem. 1964, 36, 1627--1639  
  }

\author{
Stefan Roediger, Michal Burdukiewicz
}

\seealso{
\code{\link[stats]{approx}}, \code{\link[stats]{spline}}.
}
\examples{
# Simulate a qPCR reaction with AmpSim with for 40 cycles
res <- AmpSim(cyc = c(1:40))
# Introduce a missing value in the transition between 
# the background and the exponential phase
res.NA <-res
res.NA[18, 2] <- NA
par(mfrow = c(2,2))
plot(res, xlab = "Cycles", ylab = "Fluorescence", type = "b", pch = 20, 
     main = "Simulation without missing value")
     abline(v = c(17.5, 18.5), col = "grey")
     abline(h = c(0.09, 0.14), col = "grey")

res.NA.linear <- fixNA(res.NA[, 1], res.NA[, 2], spline = FALSE, 
		       verbose = FALSE)
plot(res.NA.linear, xlab = "Cycles", ylab = "Fluorescence", type = "b", 
     pch = 20, main = "Simulation with linear imputed\n NA 
     value during transition")
     abline(v = c(17.5, 18.5), col = "grey")
     abline(h = c(0.09, 0.14), col = "grey")

plot(res.NA, xlab = "Cycles", ylab = "Fluorescence", type = "b", pch = 20, 
     main = "Simulation with missing\n value during transition")
     abline(v = c(17.5, 18.5), col = "grey")
     abline(h = c(0.09, 0.14), col = "grey")
res.NA.spline <- fixNA(res.NA[, 1], res.NA[, 2], spline = TRUE, 
		       verbose = FALSE)
plot(res.NA.spline, xlab = "Cycles", ylab = "Fluorescence", type = "b", 
      pch = 20, main = "Simulation with spline imputed\n NA value 
      during transition")
     abline(v = c(17.5, 18.5), col = "grey")
     abline(h = c(0.09, 0.14), col = "grey")     
par(mfrow = c(1,1))
}
% Add one or more standard keywords, see file 'KEYWORDS' in the
% R documentation directory.
\keyword{ spline }
\keyword{ isothermal }
\keyword{ linear }
\keyword{ amplification }