braidrm (version 1.0.3)

calcBraidConfInt: Generic BRAID confidence intervals

Description

Generates confidence intervals on derived BRAID response surface values

Usage

calcBraidConfInt(bfit, parfunc, civals = NULL)

Value

An n-by-3 array, where n is the length of the output produced by parfunc. The first column is the lower bound of the confidence interval; the second column is the derived value for the best fit coefficients; and the third column is the upper bound of the confidence intervals. Note that is possible for the lower bound of the confidence interval to lie above the central value, or for the upper bound to lie below it; though this is only likely to occur in the case of a poorly determined fit.

Arguments

bfit

A BRAID fit object of class braidrm which contains a full set of bootstrapped response surface coefficients

parfunc

A function that takes a full-length BRAID parameter vector as an input and gives a single numeric value or numeric vector as an output. If the function produces a vector, it must produce the same length vector for all inputs

civals

If given, the lower and upper quantile values at which the confidence intervals are set. Defaults to the ciLevs parameter of the bootstrapped BRAID fit

Details

In come cases, it is desirable to estimate a confidence interval on a value derived from or dependent on a BRAID surface model that is not a parameter of the model itself. For example, one might want a confidence interval on a given index of achievable efficacy value, or the predicted effect at a certain set of dose pairs. This function replicates confidence interval calculations on any such derive values

Examples

bfit <- braidrm(measure ~ concA + concB, synergisticExample, getCIs=TRUE)

calcBraidConfInt(bfit, function(p) evalBraidModel(10, 10, p))
calcBraidConfInt(bfit, function(p) estimateIAE(p, c(0.5, 0.9), c(10, 10)))