profile.discrim {sensR}R Documentation

Profile likelihood methods for discrim objects.

Description

Computes the (normalized or relative) profile likelihood for the parameters of a discrimination test, plots the normalized profile likelihood and computes profile likelihood confidence intervals.

Usage

## S3 method for class 'discrim':
profile(fitted, min = 0, max = 3, numpts = 50, ...)

## S3 method for class 'profile.discrim':
plot(x, level = c(0.99, 0.95), fig = TRUE,
            method = "natural", n = 500, ...)

## S3 method for class 'discrim':
confint(object, parm, level = 0.95, ...) 

Arguments

fitted a discrim object.
x a profile.discrim object.
object a discrim object.
parm currently not used.
min the minimum delta for which to do the profiling. By default set to 0, which for numerical stability is change internally to 1e-4.
max the maximum delta beyond the MLE for which to do the profiling.
numpts control parameter: At how many points should the profile likelihood be evaluated?
method the type of spline to be used in approximating the profile likelhood curve (trace)—se spline for details.
n the number of spline interpolations to use in plotting the profile likelihood curve (trace).
level for plot: At which levels to include horizontal lines to indicate confidence levels in plots of the normalized profile likelihoods. For confint: at which level to compute the confidence interval.
fig logical: Should the normalized profile likelihoods be plotted?
... For plot: additional arguments to plot. For confint: additional arguments to confint.glm in package MASS. For profile: additional arguments to glm.

Value

For profile: An object of class "profile.discrim", "data.frame"—a data.frame with two columns giving the value of the parameter and the corresponding value of the profile likelihood.
For plot: An object of class "nProfile.discrim", "data.frame"—the data.frame from the profile-object with an extra columns containing the normalized profile likelihood.
For confint:
A 2x2 matrix with columns named "lower", "upper" giving the lower and upper (1 - alpha)% confidence interval for the parameters named in the rows.

Author(s)

Rune Haubo B Christensen and Per Bruun Brockhoff

References

Brockhoff, P.B. and Christensen R.H.B.(2008). Thurstonian models for sensory discrimination tests as generalized linear models. Manuscript for Food Quality and Preference.

Examples

## 7 success out of 10 samples in a duo-trio experiment:
dd <- discrim(7, 10, "duotrio")
plot(profile(dd))
confint(dd)
points(confint(dd), rep(.1465, 2), pch = 3)


[Package sensR version 1.0.0 Index]