rm(list=ls(all=TRUE))
detach(PeptideDat)

PeptideDat <- read.table("/Users/tmcmurry/Documents/Consulting/Bioinformatics Project/simplelarge_commas_removed.csv", header=TRUE, sep=",", na.strings="Err:521")
attach(PeptideDat)
						
conflevel <- 0.95		#The confidence level for the intervals.  Typical values are .90, .95, .99
						#.95 means that roughly 95 percent of all intervals constructed this
						#way will contain the true mean
						#The closer this number gets to 1, the bigger the intervals get
						
residlim <-.3			#Y-scale for the residual plots

sdmatrix <- matrix(0,nrow=3, ncol=20)

pepNames <- c("PKT","ADA", "MHR", "RRL", "FHD", "TST", "ADE")
sgnNames <- c("POS", "NEG")
matNames <- levels(Matrix)
colors <- character(length(Matrix))
colors[Matrix=="CHCA"] <- "red" 	#These allow one to choose the 
colors[Matrix=="DHB"]  <- "blue"	#plotting colors for the 
colors[Matrix=="THAP"] <- "orange"	#various matrices

#The plot is a 2x7 grid of subplots.  The top row is the positives
#the second is the negatives.  The columns are the peptides in 
#alphabetical order.  Each subplot shows data for every matrix on
#which the peptide was tested.

par(mfrow = c(2,7), mar = c(3,2,2,0))

for (sgn in 1:2 )
{	
	for (pep in 1:7) 
	{

		plot(0,0,xlim=c(0,1),ylim=c(0,1),xlab="",ylab="",main=pep, type="n")		
		for (q in 1:3)
		{
		mat=matNames[q]
			#First select the right x's and y's
			#separating by matrix
		xsub <- x[Peptide == pepNames[pep] & 
#			Scientist == Sci & Experiment == Expt &
			pos_or_neg == sgnNames[sgn]&
			Matrix == mat]/100
		ysub <- y[Peptide == pepNames[pep] & 
#			Scientist == Sci & 
#			Experiment == Expt &
			pos_or_neg == sgnNames[sgn]&
			Matrix == mat]/100
			
		xsub <- xsub[!is.na(ysub)]
		ysub <- ysub[!is.na(ysub)]

		#Assign colors based on the matrix
		plotcolors <- colors[Peptide == pepNames[pep] & 
#			Scientist == Sci & Experiment == Expt & 
			pos_or_neg == sgnNames[sgn] &
			Matrix == mat]
		

		#This section deals with averages and standard deviations
		byknowns <- split(ysub,xsub)

		devs <- double(length(byknowns))
		for (i in 1:length(byknowns))
		{
			devs[i] <- sd(byknowns[[i]],na.rm=T)
		}
				
		aves <- double(length(byknowns))
		for (i in 1:length(byknowns))
		{
			aves[i] <- mean(byknowns[[i]],na.rm=T)
		}
		
		npoints <- integer(length(byknowns))
		for (i in 1:length(byknowns))
		{
			npoints[i] <- sum(!is.na(byknowns[[i]]))
		}
		
		intervals<-as.numeric(levels(factor(xsub)))
	
		
		
		wts <- rep(1, length(xsub))		

		fit.nls <- nls(ysub ~ xsub/((1-xsub)*gamma + xsub),
			start = list(gamma = 1),
			weights = wts)
			gam <- coef(fit.nls)[[1]]
	
		fittedxs <- seq(0,1,.01)
		fittedys <- fittedxs/((1-fittedxs)*gam + fittedxs)
		
		lines(fittedxs, fittedys, col=plotcolors[1])
#		abline(0,1)

		
		#Plot the confidence intervals
		for (j in 1:length(byknowns))
		{
			tmult <- -qt( (1-conflevel)/2, df = (npoints[j]-1) )/sqrt(npoints[j])
			arrows(intervals[j],aves[j]+tmult*devs[j],intervals[j],aves[j]-tmult*devs[j],
				length=.05, angle=90,code=3,col=plotcolors[1])
		}
		
		}
	
	}
}























quartz()
par(mfrow = c(2,7), mar = c(3,2,2,0))

for (sgn in 1:2 )
{	
	for (pep in 1:7) 
	{

		plot(0,0,xlim=c(0,1),ylim=c(-residlim,residlim),xlab="",ylab="",main=pep, type="n")		
		for (q in 1:3)
		{
		mat=matNames[q]
			#First select the right x's and y's
			#separating by matrix
		xsub <- x[Peptide == pepNames[pep] & 
#			Scientist == Sci & Experiment == Expt &
			pos_or_neg == sgnNames[sgn]&
			Matrix == mat]/100
		ysub <- y[Peptide == pepNames[pep] & 
#			Scientist == Sci & Experiment == Expt &
			pos_or_neg == sgnNames[sgn]&
			Matrix == mat]/100
			
		xsub <- xsub[!is.na(ysub)]
		ysub <- ysub[!is.na(ysub)]

		#Assign colors based on the matrix
		plotcolors <- colors[Peptide == pepNames[pep] & 
#			Scientist == Sci & Experiment == Expt & 
			pos_or_neg == sgnNames[sgn] &
			Matrix == mat]
		

		#This section deals with averages and standard deviations
		byknowns<-split(ysub,xsub)

		devs <- double(length(byknowns))
		for (i in 1:length(byknowns))
		{
			devs[i] <- sd(byknowns[[i]],na.rm=T)
		}
				
		aves <- double(length(byknowns))
		for (i in 1:length(byknowns))
		{
			aves[i] <- mean(byknowns[[i]],na.rm=T)
		}
		
		npoints <- integer(length(byknowns))
		for (i in 1:length(byknowns))
		{
			npoints[i] <- sum(!is.na(byknowns[[i]]))
		}
		
		intervals<-as.numeric(levels(factor(xsub)))
	
		
		
		#this is the modeling section done by Tim
		wts <- rep(1, length(xsub))		

		fit.nls <- nls(ysub ~ xsub/((1-xsub)*gamma + xsub),
			start = list(gamma = 1),
			weights = wts)
			gam <- coef(fit.nls)[[1]]
		
		points(xsub, residuals(fit.nls), col=plotcolors[1])
		abline(0,0)
		
		}
	
	}
}