DRtrace: Constructor functions and class-checking functions for...

In cir: Centered Isotonic Regression and Dose-Response Utilities

Constructor functions and class-checking functions for DRtrace and doseResponse classes

Description

Functions to create and sanity-check objects of the DRtrace (dose-response experiment trace/trajectory) and doseResponse (dose-response raw summary) classes. Note that the latter inherits from the former, purely for programming-convenience reasons.

Usage

is.DRtrace(dr)

is.doseResponse(dr)

DRtrace(y, x = NULL, cohort = NULL, noyes = FALSE, ...)

doseResponse(y, x = NULL, wt = rep(1, length(y)), noyes = FALSE, ...)

Arguments

dr	the object being checked
y, x	see Details.
cohort	(DRtrace only) specify each observation's cohorts, if there were cohorts. If all cohorts were the same size, then you can specify the size as a single number. If there were no cohorts, code will default this variable to 1:n
noyes	logical, in case of a 2-column input is the 1st column 'no'? Default FALSE, meaning the 1st column is 'yes'.
...	parameters passed on to DRtrace(), or ignored.
wt	(doseResponse only) the weights associated with each x value; usually the sample size or similar.

Details

The input argument y can include the entire information, or as little as the y vector of responses (for a DRtrace object) or response rates (doseResponse). When including the entire information, it has to be a data frame with at least y (both y and x for DRtrace), or a two-column matrix with 'yes' and 'no' responses (assumed in this order, but can be the reverse with noyes=TRUE). In this case the doses x can be provided as a separate vector, or as the matrix row names. doseResponse will return an error if there are any duplicates in x.

Even though both DRtrace and doseResponse accept two-column yes/no matrix input, the interpretation is different. For the former, this form of input is intended mostly to enable shorthand input when the experiment was run in cohorts. Each row represents a cohort's results, and rows must be in the order the experiment was run. For the latter, the yes-no table is a summary tabulation of responses and is treated accordingly, including rearrangement of rows to increasing x.

Value

For constructor functions, the relevant object. For checking functions, a logical value indicating whether the object meets class definition.

Author(s)

Assaf P. Oron <assaf.oron.at.gmail.com>

See Also

cirPAVA, plot.doseResponse,plot.DRtrace

Examples

## Summary of raw data from the notorious Neuenschwander et al. (Stat. Med., 2008) trial
## Note the use of the 'cohort' argument to specify the cohort order
neundatTrace = DRtrace(x = c(rep(1:4,each=4), 7, 7, rep(6,9)),
                       y = c(rep(0,16), 1,1, rep(c(0,0,1),2), 0,0,0), 
                       cohort = rep(1:8, c(4,4,4,4, 2, 3,3,3)) )
par(mar=c(3,3,3,1), mgp=c(2,.5,0), tcl=-0.25)
layout(t(1:2))
plot(neundatTrace ,main="N. et al. (2008) Cohort Trace", xlab = 'Cohort', 
          ylab="Ordinal Dose Level" ,cex.main=1.5)

## Same data, in 'doseResponse' format with actual doses rather than dose levels
neundatDose = doseResponse(x=c(1,2.5,5,10,20,25), y = c(rep(0,4),2/9,1), wt = c(3,4,5,4,9,2) )
plot(neundatDose ,main="N. et al. (2008) Final Dose-Toxicity", ylim=c(0,1),
	xlab="Dose (mg/sq.m./wk)", ylab="Toxicity Response Curve (F)", cex.main=1.5)
## We can also convert the DRtrace object to doseResponse...
neundatLevel = doseResponse(neundatTrace)

### Now plotting the former, vs. IR/CIR estimates
neunCIR0 = cirPAVA(neundatDose,full=TRUE, adaptiveShrink = TRUE, target = 0.3)
lines(neunCIR0$shrinkage$x, neunCIR0$shrinkage$y, type='b' ,pch=19)
legend(1,1, pch=c(4,19), legend=c('Observations', 'CIR w/bias corr.'), bty='n')

cir documentation built on April 27, 2023, 9:05 a.m.