logistf (version 1.26.0)

profile.logistf: Compute Profile Penalized Likelihood

Description

Evaluates the profile penalized likelihood of a variable based on a logistf model fit

Usage

# S3 method for logistf
profile(
  fitted,
  which,
  variable,
  steps = 100,
  pitch = 0.05,
  limits,
  alpha = 0.05,
  firth = TRUE,
  legends = TRUE,
  control,
  plcontrol,
  ...
)

Value

An object of class logistf.profile with the following items:

beta

Parameter values at which likelihood was evaluated

stdbeta

Parameter values divided by standard error

profile

profile likelihood, standardized to 0 at maximum of likelihood. The values in profile are given as minus \(\chi^2\)

loglik

Unstandardized profile likelihood

signed.root

signed root (z) of \(\chi^2\) values (negative for values below the maximum likelihood estimate, positive for values above the maximum likelihood estimate)

cdf

profile likelihood expressed as cumulative distribution function, obtained as \(\Phi(z)\), where \(\Phi\) denotes the standard normal distribution function.

Arguments

fitted

An object fitted by logistf

which

A righthand formula to specify the variable for which the profile should be evaluated, e.g., which=~X).

variable

Alternatively to which, a variable name can be given, e.g., variable="X"

steps

Number of steps in evaluating the profile likelihood

pitch

Alternatively to steps, one may specify the step width in multiples of standard errors

limits

Lower and upper limits of parameter values at which profile likelihood is to be evaluated

alpha

The significance level (1-\(\alpha\) the confidence level, 0.05 as default).

firth

Use of Firth's penalized maximum likelihood (firth=TRUE, default) or the standard maximum likelihood method (firth=FALSE) for the logistic regression.

legends

legends to be included in the optional plot

control

Controls Newton-Raphson iteration. Default is control= logistf.control(maxstep, maxit, maxhs, lconv, gconv, xconv)

plcontrol

Controls Newton-Raphson iteration for the estimation of the profile likelihood confidence intervals. Default is plcontrol= logistpl.control(maxstep, maxit, maxhs, lconv, xconv, ortho, pr)

...

Further arguments to be passed.

References

Heinze G, Ploner M, Beyea J (2013). Confidence intervals after multiple imputation: combining profile likelihood information from logistic regressions. Statistics in Medicine, to appear.

Examples

data(sex2)
fit<-logistf(case ~ age+oc+vic+vicl+vis+dia, data=sex2)
plot(profile(fit,variable="dia"))
plot(profile(fit,variable="dia"), "cdf")
plot(profile(fit,variable="dia"), "density")