profLik: Profile likelihood for coefficients in a 'coxph' model

View source: R/profLik.R

profLikR Documentation

Profile likelihood for coefficients in a coxph model


Profile likelihood for coefficients in a coxph model


profLik(x, CI = 0.95, interval = 50, mult = c(0.1, 2), devNew = TRUE, ...)



A coxph model.


Confidence Interval.


Number of points over which to evaluate coefficient.


Multiplier. Coefficent will be multiplied by lower and upper value and evaluated across this range.


Open a new device for each plot. See


Additional parameters passed to graphics::plot.default.


Plots of range of values for coefficient in model with log-likelihoods for the model with the coefficient fixed at these values.

For each coefficient a range of possible values is chosen, given by Bhat * mult[lower] - Bhat * mult[upper]. A series of models are fit (given by interval). The coefficient is included in the model as a fixed term and the partial log-likelihood for the model is calculated.

A curve is plotted which gives the partial log-likelihood for each of these candidate values. An appropriate confidence interval (CI) is given by subtracting 1/2 the value of the appropriate quantile of a chi-squared distribution with 1 degree of freedom.

Two circles are also plotted giving the 95


One plot for each coefficient in the model.


Example is from: T&G. Section 3.4.1, pg 57.


data("pbc", package="survival")
c1 <- coxph(formula = Surv(time, status == 2) ~ age + edema + log(bili) +
                      log(albumin) + log(protime), data = pbc)
profLik(c1, col="red")

survMisc documentation built on April 7, 2022, 5:06 p.m.