predictBias {nnNorm}R Documentation

Computes the bias estimate provided a neural network color distortion model

Description

Given a neural network color distorsion model and an intensity value, this function computes de estimate of the bias of log ratios, M. The color distorsion model should be obtained with maNormNN.

Usage

predictBias(Avals,Models,slide,pT) 

Arguments

Avals An array of average log intensities for which the bias will be estimated. All the values in array must be inside the validity ranges of the model, i.e. form Models$lims[1,pT,slide] to Models$lims[3,pT,slide]
Models The models component of the list returned by the function maNormNN.
slide The slide number for which prediction is desired.
pT The print tip number for which prediction is desired.

Details

This function is used to make interpolations usings the color distortion models saved during the normalization effectuated with the function maNormNN.

Value

An array of the same length as Avals with the coresponding bias estimates.

Author(s)

Tarca, A.L.

References

A. L. Tarca, J. E. K. Cooke, and J. Mackay. Robust neural networks approach for spatial and intensity dependent normalization of cDNA data. Bioinformatics. 2004,submitted.

See Also

maNormNN

Examples

 
data(swirl)
#normalize swirl
sNNr<-maNormNN(swirl,binWidth=maNsc(swirl),binHeight=maNsr(swirl),save.models=TRUE,robust=TRUE)

#retrive original M-A values for slide 4 and print tip 3
s<-4;pt<-3;
MM<-maM(swirl[maPrintTip(swirl)==pt,s])
AA<-maA(swirl[maPrintTip(swirl)==pt,s])

#generate a series of A values in the validity range of the model for slide s and print tip pt
A<-seq(sNNr$models$lims[1,pt,s],sNNr$models$lims[3,pt,s],length=100)

#do the plots
if(interactive()){x11()}
plot(AA,MM,pch=20,xlab="A",ylab="M",main=paste("Slide=",s," Print tip=",pt)) #raw data
lines(A,predictBias(A,sNNr$models,s,pt),col="red",lwd=2) #robust NN

#for ilustration add the loess normalization curve as computed in marray package
lo<-loess(MM~AA,span=0.4,degree=1,family="symmetric",control=loess.control(trace.hat="approximate",iterations=5,surface="direct"))
lines(A,predict(lo,A),col= "green",lty="longdash",lwd=2) #loess


[Package nnNorm version 1.0.1 Index]