Bootstrap test for testing dose response curves for similarity concerning the maximum absolute deviation

Share:

Description

Function for testing whether two dose response curves can be assumed as equal concerning the hypotheses

H_0: \max_{x\in\mathcal{X}} |m_1(d,θ_1)-m_2(d,θ_2)|≥q ε\ vs.\ H_1: \max_{x\in\mathcal{X}} |m_1(d,θ_1)-m_2(d,θ_2)|< ε.

See http://arxiv.org/pdf/1505.05266.pdf for details.

Arguments

data1,data2

data frame for each of the two groups containing the variables referenced in dose and resp

m1,m2

model types. Built-in models are "linlog", "linear", "quadratic", "emax", "exponential", "sigEmax", "betaMod" and "logistic"

epsilon

positive argument specifying the hypotheses of the test

B

number of bootstrap replications. If missing, default value of B is 5000

bnds1,bnds2

bounds for the non-linear model parameters. If not specified, they will be generated automatically

plot

if TRUE, a plot of the absolute difference curve of the two estimated models will be given

scal,off

fixed dose scaling/offset parameter for the Beta/ Linear in log model. If not specified, they are 1.2*max(dose) and 1 respectively

Value

A list containing the p.value, the maximum absolute difference of the models, the estimated model parameters and the number of bootstrap replications. Furthermore plots of the two models are given.

References

http://arxiv.org/pdf/1505.05266.pdf

Examples

1
2
3
4
5
6
library("DoseFinding")
library("alabama")
data(IBScovars)
male<-IBScovars[1:118,]
female<-IBScovars[119:369,]
bootstrap_test(male,female,"linear","emax",epsilon=0.35,B=300)