sc13-getConfidenceInterval: Compute Confidence Intervals for a Model Fit to Dilution...
In SuperCurve: RPPA Analysis Package

Description Usage Arguments Details Value Author(s) See Also Examples

This function computes confidence intervals for the estimated concentrations in a four-parameter logistic model fit to a set of dilution series in a reverse-phase protein array experiment.

getConfidenceInterval(result,
                      alpha=0.1,
                      nSim=50,
                      progmethod=NULL)

`result`	object of class `RPPAFit` representing the result of fitting a four-parameter logistic model
`alpha`	numeric scalar specifying desired significance of the confidence interval; the width of the resulting interval is 1 - alpha.
`nSim`	numeric scalar specifying number of times to resample the data in order to estimate the confidence intervals.
`progmethod`	optional function that can be used to report progress.

In order to compute the confidence intervals, the function assumes that the errors in the observed Y intensities are independent normal values, with mean centered on the estimated curve and standard deviation that varies smoothly as a function of the (log) concentration. The smooth function is estimated using loess. The residuals are resampled from this estimate and the model is refit; the confidence intervals are computed empirically as symmetrically defined quantiles of the refit parameter sets.

An object of class RPPAFit, containing updated values for the slots lower, upper, and conf.width that describe the confidence interval.

Kevin R. Coombes kcoombes@mdanderson.org, P. Roebuck proebuck@mdanderson.org

RPPAFit-class, RPPAFit

  ## Not run: 
extdata.dir <- system.file("extdata", package="SuperCurveSampleData")

txtdir <- file.path(extdata.dir, "rppaCellData")
akt <- RPPA("Akt.txt", path=txtdir)
design <- RPPADesign(akt,
                     grouping="blockSample",
                     controls=list("neg con", "pos con"))
fit.nls <- RPPAFit(akt, design, "Mean.Net")
## N.B.: this takes a while!
fit.nls <- getConfidenceInterval(fit.nls, alpha=0.10, nSim=50)
  
## End(Not run)