mcprofile (version 0.0-9)

confint-methods: Simultaneous Confidence Intervals for Multiple Contrast Profiles

Description

Calculates simultaneous confidence intervals based on mcprofile objects. Confidence limits are found by spline interpolation at a signed root deviance cutoff value. Multiplicity adjustment can be performed by Bonferroni correction or utilizing a multivariate normal or t-distribution.

Usage

## S3 method for class 'mcprofile':
confint(object, parm, level=0.95,
adjust="single-step", alternative="two.sided", quant=NULL)

Arguments

object

An object of class mcprofile, or mcprofileRatio

parm

Missing, only needed to provide a mcprofile method for confint

level

Global confidence level, default at 0.95

adjust

Multiplicity adjustment method: "none" for unadjusted intervals, "bonferroni", or "single-step" for utilizing a multivariate normal or t-distribution.

alternative

Choosing one- or two-sided intervals by "two.sided", "less", or "greater"

quant

Numeric, in not NULL this value is used as cutoff value for the signed root deviance

Value

• An object of class mcpconfint

Details

Confidence limits are found by spline interpolation using the splines in the bsplines slot of the mcprofile, or mcprofileRatio object. Cutoff values are calculated as quantiles of a normal distribution, or t-distribution if the df slot is not empty (gaussian families). For "single-step" adjustment the correlation structure of a multivariate normal distribution is estimated by standardizing the variance-covariance matrix of the linear combinations of parameters. This vcov matrix is obtained by multiplying the original vcov matrix with the prespecified contrast matrix from both sides.

See Also

mcpcalc, mcpcalcRatio, confint.glht, sci.ratio