DoseFinding (version 0.6-3)

bootMCPMod: Evaluate precision of dose estimate by nonparametric bootstrapping

Description

Applies nonparametric bootstrapping to evaluate the precision of the dose-estimate, the dose-response estimate and the model selection of a MCPMod object. Bootstrapping is done stratified, so that we resample within each dose group.

Usage

bootMCPMod(x, nSim = 1000, fitControl = list(), start = NULL,
             seed = 100, calcDR = FALSE, ...)

Arguments

x

A MCPMod object, with a calculated critical value (see critV argument in MCPMod function)

nSim

Number of bootstrap replications

fitControl

List of control parameters for nonlinear fitting (see fit.control for details), passed down to MCPMod function

start

List of starting values for nls, passed down to MCPMod function during bootstrapping

seed

Random seed (to be able to reproduce results)

calcDR

Logical values indicating, whether bootstrap dose-response estimates should be calculated and saved in the bootMCPMod object.

...

Additional arguments passed to the predict.MCPMod function, in case calcDR = TRUE, see predict.MCPMod for details.

Value

• An object of class bootMCPMod, ie. a list with entries doseEst, model and (if calcDR is TRUE) doseResp, containing the bootstrap replications. In case of model averaging doseEstMods contains the bootstrap replications of the dose estimates of the different significant models. Additionally the original fitted MCPMod object is included.

See Also

MCPMod

Examples

data(biom)
models <- list(linear = NULL, linlog = NULL, emax = 0.2,
            exponential = c(0.279,0.15), quadratic = c(-0.854,-1))
dfe <- MCPMod(resp ~ dose, biom, models, alpha = 0.05, doseEstPar = 0.05,
           selModel = "maxT", doseEst = "MED2old", clinRel = 0.4, off = 1,
           critV = TRUE, optimizer = "bndnls")
btMED <- bootMCPMod(dfe, nSim = 1000)
btMED
## more information on quantiles of dose-estimate
quantile(btMED$doseEst, c(0.05,0.25,0.5,0.75,0.95), na.rm=TRUE)
## plot density estimate of dose-estimate
dens <- density(btMED$doseEst, from = 0, to = 1, na.rm=TRUE)
plot(dens, main = "Bootstrap MED Replications")
## summarize selected models
table(btMED$model)/1000

## more time consuming
## now use model averaging and estimate ED50
data(IBScovars)
models <- list(emax = 0.2, quadratic = -0.2, linlog = NULL)
dfe <- MCPMod(resp ~ dose, IBScovars, models, addCovars = ~gender,
             alpha = 0.05, critV = TRUE, selModel = "aveAIC", doseEst = "ED",
             clinRel = 0.25, off = 1, optimizer = "bndnls")
## now also investigate uncertainty in DR estimate
## predict at sq
sq <- seq(0, 4, length = 101)
## Note: type = "EffectCurve", doseSeq = sq are passed to predict.MCPMod
btED <- bootMCPMod(dfe, nSim = 1000, calcDR = TRUE,
                    type = "EffectCurve", doseSeq = sq)
btED
## display information on ED50 estimate
dens <- density(btED$doseEst, from = 0, to = 4, na.rm=TRUE)
plot(dens, main = "Bootstrap MED Replications")
## plot uncertainty quantiles for dose-response curve
res <- apply(btED$doseResp, 2, quantile,
             prob = c(0.1,0.25, 0.5,0.75, 0.9), na.rm=T)
plot(sq, res[3,], ylim = c(0, 0.6), type = "l",
     xlab = "Dose", ylab = "Effect Curve")
polygon(c(sq,rev(sq)), c(res[1,], rev(res[5,])), col = "lightgrey")
polygon(c(sq,rev(sq)), c(res[2,], rev(res[4,])), col = "grey")
lines(sq, res[3,], type = "l")