preProcess {TIMP}

Performs preprocessing on data stored as an objects of class dat.

Description

Performs data sampling, selection, baseline correction, scaling, and data correction on an object of class dat.

Usage

preProcess(data, sample = 1, sample_time = 1, sample_lambda = 1,
    sel_time = vector(), sel_lambda = vector(), baselinetime = vector(),
    baselinelambda = vector(), scalx = NULL, scalx2 = NULL,
    sel_lambda_ab = vector(), sel_time_ab = vector(), rm_x2=vector(),
    rm_x = vector(), svdResid = list(), numV = 0, sel_special = list(),
    doubleDiff = FALSE, doubleDiffFile = "doubleDiff.txt")

Arguments

data

Object of class dat

sample

integer describing sampling interval to take in both time and x2; e.g., sample=2 will sample every 2nd time and every 2nd point in x2.

sample_time

integer describing sampling interval in time; e.g., sample_time=2 will sample every 2nd element of the time vector.

sample_lambda

integer describing sampling interval in x2; e.g., sample_lambda=2 will sample every 2nd element in the x2 vector.

sel_time

vector of length 2 describing the first and last time index of data to select; e.g., sel_time=c(5,120) will select data at times indexed 5-120.

sel_lambda

vector of length 2 describing the first and last x2 index of data to select; e.g., sel_lambda=c(5,120) will select data at x2 indexed 5-120.

baselinetime

a vector of form c(timeIndexmin, timeIndexmax, lambdaIndexmin, lambdaIndexmax). The average of data between x2 indexes lambdaIndexmin and lambdaIndexmax is subtracted from data with time index between timeIndexmin and timeIndexmax.

baselinelambda

a vector of form c(timeIndexmin, timeIndexmax, lambdaIndexmin, lambdaIndexmax). The average of data between time indexes timeIndexmin and timeIndexmax is subtracted from data with x2 index between lambdaIndexmin and lambdaIndexmax.

scalx

numeric by which to linearly scale the x axis (which often represents time), so that newx = oldx * scalx

scalx2

vector of length 2 by which to linearly scale the x2 axis, so that newx2 = oldx2 * scalx2[1] + scalx2[2]

sel_lambda_ab

vector of length 2 describing the absolute values (e.g., wavelengths, wavenumbers, etc.) between which data should be selected. e.g., sel_lambda_ab = c(400, 600) will select data associated with x2 values between 400 and 600.

sel_time_ab

vector of length 2 describing the absolute times between which data should be selected. e.g., sel_time_ab = c(50, 5000) will select data associated with time values between 50 and 5000 picoseconds.

rm_x2

vector of x2 indices to remove from the data

rm_x

vector of x indices to remove from the data

svdResid

list returned from the getResid function, containing residuals to be used in data correction.

numV

numeric specifying how many singular vectors to use in data correction. Maximum is five.

sel_special

list of lists specifying x indices to remove from individual wavelength ranges, e.g., sel_special = list(c(400,600,10,12),c(600,800,11,13)) indicates that between wavelength 400 and 600, time indices between 10 and 12 should be removed from the data, and between wavelengths 600 and 800, time indices between 11 and 13 should be removed from the data. Note that the number of time indices to remove should be the same in each wavelength interval specified. Also note that the time vector associated with the data after the first set of indices is removed will be associated with the resulting dataset.

doubleDiff

logical indicating whether the data should be converted to represent differences between times.

doubleDiffFile

character string indicating the file name of time difference data to create in the case that doubleDiff=TRUE.

Value

object of class dat.

Author(s)

Katharine M. Mullen, Ivo H. M. van Stokkum

See Also

readData, getResid

Examples

##############################
## READ DATA
##############################

data("target")

##############################
## PREPROCESS DATA
##############################

# select certain wavelengths for modeling

C1_1 <- preProcess(data = C1, baselinelambda = c(1,12,1,32) )
C1_1 <- preProcess(data = C1_1, sel_lambda = c(8, 27))
C1_1 <- preProcess(data = C1_1, rm_x = c(40, 41, 101, 116))
C1_1 <- preProcess(data = C1_1, sel_time_ab = c(-10, 100000))

C2_1 <- preProcess(data = C2, sel_lambda = c(2, 32))
C2_1 <- preProcess(data = C2_1, baselinelambda = c(1,12,1,32) )
C2_1 <- preProcess(data = C2_1, sel_time_ab = c(-10, 100000))

C3_1 <- preProcess(data = C3, sel_lambda = c(1, 25))
C3_1 <- preProcess(data = C3_1, baselinelambda = c(1,12,1,32) )



##############################
## SPECIFY K Matrix and J vector
##############################


## initialize 2 7x7 arrays to 0

delK <- array(0, dim=c(7,7,2))

## the matrix is indexed:
## delK[ ROW K MATRIX, COL K MATRIX, matrix number]

## in the first matrix, put the index of compartments
## that are non-zero
## the transfer rate of the compartment is governed by
## kinpar[index]

delK[1,1,1] <- 4
delK[5,1,1] <- 1
delK[2,2,1] <- 4
delK[5,2,1] <- 2
delK[3,3,1] <- 4
delK[5,3,1] <- 3
delK[4,4,1] <- 4
delK[6,5,1] <- 5
delK[7,6,1] <- 6
delK[7,7,1] <- 7

## print out the resulting array to make sure it's right

delK

jvector <- c(.48443195136500550341, .28740782363398824522,
.13749071230100625137, 0.9066953510E-01, 0, 0, 0)

datalist <- list(C1, C2, C3)
## for plotting selected traces, get a vector of all the wavenumbers
allx2 <- vector()
for(i in 1:length(datalist))
	allx2 <- append(allx2,datalist[[i]]@x2)
allx2 <- sort(unique(allx2))

##############################
## SPECIFY INITIAL MODEL
## note that low is the larger wavenumber in the clpequ spec!
##############################

model1 <- initModel(mod_type = "kin",
kinpar=c( 0.13698630, 0.3448275849E-01, 0.1020408142E-01, 0.2941176528E-02,
0.17000,  0.015,  0.1074082902E-03),
fixed = list(prel = 1:6, clpequ=1:3, kinpar=1:7, irfpar=1, parmu=1),
irfpar=c(0.4211619198, 0.6299000233E-01),
prelspec = list(
list(what1="kinpar", ind1=1, what2 = "kinpar", ind2=4,
start=c(-1,0.1369863003)),
list(what1="kinpar", ind1=2, what2 = "kinpar", ind2=4,
start=c(-1,0.3448275849E-01)),
list(what1="kinpar", ind1=3, what2 = "kinpar", ind2=4,
start=c(-1,0.1020408142E-01))
),
parmu = list(c(-0.1411073953)),
lambdac = 1290,
kmat = delK,
jvec = jvector,
positivepar="kinpar",
weightpar=list( c(-20,1.4,1,2000,.2)),
clpequspec =list(
list(to=2, from=1, low=100, high=10000),
list(to=3, from=1, low=100, high=10000),
list(to=4, from=1, low=100, high=10000)),
clpequ = c(1,1,1),
cohspec = list( type = "irf"))

##############################
## GET RESID
## same format as call to fitModel, but does not plot
##############################

serResid <- getResid(list(C1_1, C2_1, C3_1), list(model1),
modeldiffs = list(thresh =  0.00005,
dscal = list(list(to=2,from=1,value=4),
list(to=3,from=1,value=0.8000000119)),
free = list(
list(what="irfpar", ind=1, start= c(0.1231127158), dataset=2),
list(what="parmu", ind=c(1,1), start= c(0.1219962388), dataset=2),
list(what="irfpar", ind=1, start= c(0.3724052608), dataset=3),
list(what="parmu", ind=c(1,1), start= c(0.8844097704E-01), dataset=3)),
change  = list(
list(what="fixed", spec=list(clpequ=1:3, kinpar=1:7, irfpar=1:2,
parmu=1, drel = 1, prel=1:6), dataset=2:3))),
opt=kinopt(iter=0, title="Cosimo Spectra, Not Normalized, with Error",
stderrclp=TRUE, kinspecerr=TRUE, writespec = TRUE,
plotkinspec = TRUE,plotcohcolspec=FALSE,
selectedtraces = seq(1, length(allx2), by=2),
specinterpol = TRUE, specinterpolpoints=FALSE,
divdrel=TRUE, xlab="wavenumber",writeclperr = TRUE,
makeps = "err", linrange = 1, superimpose=1:3))

##############################
## MAKE CORRECTED DATASETS USING RESID INFO
##############################

C1_3 <- preProcess(data = C1_1, svdResid = serResid[[1]], numV = 2)
C2_3 <- preProcess(data = C2_1, svdResid = serResid[[2]], numV = 2)
C3_3 <- preProcess(data = C3_1, svdResid = serResid[[3]], numV = 2)

##############################
## FIT MODEL
##############################

serRes<-fitModel(list(C1_3, C2_3, C3_3), list(model1),
modeldiffs = list(thresh =  0.00005,
dscal = list(list(to=2,from=1,value=4),
list(to=3,from=1,value=0.8000000119)),
free = list(
list(what="irfpar", ind=1, start= c(0.1231127158), dataset=2),
list(what="parmu", ind=c(1,1), start= c(0.1219962388), dataset=2),
list(what="irfpar", ind=1, start= c(0.3724052608), dataset=3),
list(what="parmu", ind=c(1,1), start= c(0.8844097704E-01), dataset=3)),
change  = list(
list(what="fixed", spec=list(clpequ=1:3, kinpar=1:7, irfpar=1:2,
parmu=1, drel = 1, prel=1:6), dataset=2:3))),
opt=kinopt(iter=0, title="Cosimo Spectra, Not Normalized, with Error",
stderrclp=TRUE, kinspecerr=TRUE, writespec = TRUE,
plotkinspec = TRUE,plotcohcolspec=FALSE, writerawcon = TRUE,
selectedtraces = seq(1, length(allx2), by=2),
specinterpol = TRUE, specinterpolpoints=FALSE,
divdrel=TRUE, xlab="wavenumber",writeclperr = TRUE,
makeps = "h20", linrange = 1, superimpose=1:3))

 # end donttest
##############################
## CLEANUP GENERATED FILES
##############################
# This removes the files that were generated in this example
# (do not run this code if you wish to inspect the output)
file_list_cleanup = c('h20_paramEst.txt', 'h20_rawconcen_dataset_1.txt',
  'h20_rawconcen_dataset_2.txt', 'h20_rawconcen_dataset_3.txt',
  'h20_spec_dataset_1.txt', 'h20_spec_dataset_2.txt',
  'h20_spec_dataset_3.txt', 'h20_std_err_clp_1.txt',
  'h20_std_err_clp_2.txt', 'h20_std_err_clp_3.txt',
  'err_paramEst.txt', 'err_spec_dataset_1.txt', 'err_spec_dataset_2.txt',
  'err_spec_dataset_3.txt', 'err_std_err_clp_1.txt',
  'err_std_err_clp_2.txt', 'err_std_err_clp_3.txt',
  Sys.glob("*paramEst.txt"), Sys.glob("*.ps"), Sys.glob("Rplots*.pdf"))

# Iterate over the files and delete them if they exist
for (f in file_list_cleanup) {
  if (file.exists(f)) {
    unlink(f)
  }
}