getResults: Functions to print and return parts of the object returned by the fitting routines.

Description

Functions to print and return parts of the object returned by fitModel. onls returns the output of the nls function. sumonls returns the result of calling summary on onls function. parEst returns a summary of model parameter estimates. The remaining functions return lists representing various aspects of the results returned by the function fitModel.

Usage

onls(result)
sumnls(result)
parEst(result, param = "", dataset = NA, verbose = TRUE, file="",
 stderr=TRUE)
getXList(result, group = vector(), file="")
getCLPList(result, getclperr = FALSE, file="")
getX(result, group = vector(), dataset=1, file="", lreturnA = FALSE, lreturnC = FALSE)
getC(result, dataset=1, file="")
getCLP(result, getclperr = FALSE, dataset=1, file="")
getDAS(result, getclperr = FALSE, dataset=1, file="")
getData(result, dataset = 1, weighted = FALSE)
getResiduals(result, dataset = 1)
getSVDResiduals(result, numsing = 2, dataset = 1)
getTraces(result, dataset = 1, file="")
getdim1(result, dataset = 1)
getdim2(result, dataset = 1)

Value

sumnls returns an object of class "summary.nls".

onls returns an object of class "nls".

parEst returns an object of class "list" representing the parameter estimates and the standard errors if stderr=TRUE

and they have been calculated.

getXList returns a "list" of length equal to the number of datasets modeled, where each element represents the matrix determined by the nonlinear parameters (under a kinetic model, the concentrations).

getCLPList returns a "list" of length equal to the number of datasets modeled, where each element represents the matrix determined as conditionally linear parameters (under a kinetic model, the spectra).

getX returns a numeric "matrix"

that represents the matrix determined by the nonlinear parameters (under a kinetic model, the concentrations). However, in case lreturnC = TRUE it returns the C matrix, and in case lreturnA = TRUE it returns the A matrix that is used to compute the C matrix in case the kinetic model differs from parallel decays.

getC returns (under a kinetic model) a numeric "matrix"

that represents the raw matrix of concentrations of the dataset determined by the nonlinear parameters.

getDAS returns (under a kinetic model) a numeric "matrix"

that represents the Decay Associated Spectra (DAS).

getCLPList returns a numeric "matrix"

that represents the matrix determined as conditionally linear parameters (under a kinetic model, the spectra).

getSVDData

returns a "list" of length 3 with named elements

values, left and right, where values

contains the singular values, left contains numsing

left singular vectors, and right contains numsing

right singular vectors, all of the unweighted data. The number of singular vectors returned is determined by numsing.

getData returns the dataset specified by the argument dataset (weighted data in the case that weighted=TRUE) as a "matrix"

getResiduals returns a "matrix" of residuals for the dataset with index given by the argument dataset; the matrix returned has the dimension of the dataset itself.

getSVDResiduals

returns a "list" of length 3 with named elements

values, left and right, where values

contains the singular values, left contains numsing

left singular vectors, and right contains numsing

right singular vectors, all of the residuals. The number of singular vectors returned is determined by numsing.

getTraces returns a "matrix" of model estimates for the dataset with index given by the argument dataset; the matrix returned has the dimension of the dataset itself.

getdim1 returns a "vector" of

x values in the dataset (times for kinetic models).

getdim2 returns a "vector" of

x2 values (wavelengths for kinetic models).

Arguments

result: return value of fitModel
param: character vector of the particular parameters to return; if param=="" then all parameters are given.
dataset: index of the dataset from which to return results; by default dataset=NA in which case results from all datasets are returned
verbose: logical that defaults to TRUE that determines whether parEst just returns a list invisibly or prints as well.
getclperr: logical that defaults to FALSE that determines whether a list containing the standard error estimates associated with the conditionally linear parameters, as opposed to the conditionally linear parameters themselves
numsing: integer that defaults to 2; determines the number of singular vectors to return
weighted: logical indicating whether to return weighted or unweighted data
lreturnA: logical indicating whether to return an A matrix instead
lreturnC: logical indicating whether to return a C matrix instead
file: character vector; if not "" writes the results to a file with name file.
group: The value at which to determine the X matrix (maybe a wavelength index, for example)
stderr: Whether to return standard error estimates on parameters, if they were calculated in fitting.

Author

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

Examples

Run this code


# \donttest{

## Example showing the addition of non-negativity constraints to
## conditionally linear parameters (here the spectra associated with
## a kinetic model)

## For the 1st simulated dataset, the constraints offer a modest improvement
## in the estimated spectra, whereas for the 2nd simulated dataset, they
## prevent a catastrophe in which the estimated components are hugely
## compensating.

##############################
## load TIMP
##############################

require(TIMP)

##############################
## set random seed for reproducability of noise
##############################

set.seed(80)

##############################
## SIMULATE DATA, noise realization 1
##############################

 dt4 <- simndecay_gen(kinpar = c(.4, .8, 2), seqmod = FALSE, tmax
     = 2, deltat = .04, specpar = list(c(25000, 3000, .01), c(22000,
     3000, .01), c(18000, 3000, .01)), lmin=350, lmax=550, deltal = 2,
     sigma=.01)

##############################
## SPECIFY INITIAL MODEL
##############################

mod1 <- initModel(mod_type = "kin", kinpar = c(.4, .8, 2),
seqmod=FALSE)

##############################
## FIT INITIAL MODEL
##############################

sT <- fitModel(list(dt4), list(mod1), opt=kinopt(iter=50, plot=FALSE))

##############################
## EXTRACT ESTIMATED SPECTRA
## these spectra have some negative values
##############################

sTcp <- getCLP(sT)

## plot the estimated spectra with the values used in
## simulation  (before adding noise) for comparison
 matplot(dt4@x2, sTcp, xlab = "wavelength (nm)", col = 2:4, type="l",
        ylab="",lty=1, main =
        paste("Estimated spectra, adding no constraints\n"))
 matplot(dt4@x2,dt4@E2, add=TRUE, type="l", col=1, lty=2)
 abline(0,0)


##############################
## FIT INITIAL MODEL
## adding constraints to non-negativity of the
## spectra via the opt option nnls=TRUE
##############################

sV <- fitModel(list(dt4), list(mod1), opt=kinopt(iter=50, nnls=TRUE,
                                        plot=FALSE))

##############################
## EXTRACT ESTIMATED SPECTRA
## these spectra have no negative values
##############################

sVcp <- getCLP(sV)

## plot the estimated spectra with the values used in
## simulation  (before adding noise) for comparison
 matplot(dt4@x2, sVcp, xlab = "wavelength (nm)", col = 2:4, type="l",
 ylab="",lty=1,
         main = paste("Estimated spectra, with non-negativity constraints\n"))
 matplot(dt4@x2,dt4@E2, add=TRUE, type="l", col=1, lty=2)
 abline(0,0)

##############################
## SIMULATE DATA, noise realization 2
##############################

 dt4_2 <- simndecay_gen(kinpar = c(.4, .8, 2), seqmod = FALSE, tmax
     = 2, deltat = .04, specpar = list(c(25000, 3000, .01), c(22000,
     3000, .01), c(18000, 3000, .01)), lmin=350, lmax=550, deltal = 2,
     sigma=.01)

##############################
## SPECIFY INITIAL MODEL
##############################

mod1 <- initModel(mod_type = "kin", kinpar = c(.4, .8, 2),
seqmod=FALSE)

##############################
## FIT INITIAL MODEL
##############################

sT <- fitModel(list(dt4_2), list(mod1), opt=kinopt(iter=50,plot=FALSE))

##############################
## EXTRACT ESTIMATED SPECTRA
## these spectra have some negative values
##############################

sTcp <- getCLP(sT)

## plot the estimated spectra with the values used in
## simulation (before adding noise) for comparison
 matplot(dt4@x2, sTcp, xlab = "wavelength (nm)", col = 2:4, type="l",
        ylab="",lty=1, main =
        paste("Estimated spectra, adding no constraints\n"))
 matplot(dt4@x2,dt4@E2, add=TRUE, type="l", col=1, lty=2)
 abline(0,0)

##############################
## FIT INITIAL MODEL
## adding constraints to non-negativity of the
## spectra via the opt option nnls=TRUE
##############################

sV <- fitModel(list(dt4_2), list(mod1), opt=kinopt(iter=50, nnls=TRUE,
                                          plot=FALSE))

##############################
## EXTRACT ESTIMATED SPECTRA
## these spectra have no negative values
##############################

sVcp <- getCLP(sV)

## plot the estimated spectra with the values used in
## simulation (before adding noise) for comparison
 matplot(dt4@x2, sVcp, xlab = "wavelength (nm)", col = 2:4, type="l",
 ylab="",lty=1,
         main = paste("Estimated spectra, with non-negativity constraints\n"))
 matplot(dt4@x2,dt4@E2, add=TRUE, type="l", col=1, lty=2)
 abline(0,0)

# } # 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(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)
  }
}