sig: Significiance tests of coefficients in a coxph model
Description
Significiance tests of coefficients in a coxph model
Usage
sig(x, ...)
## S3 method for class 'coxph':
sig(x, ...)
Arguments
x
A model of class coxph
...
Additional arguments (not implemented)
Value
A data.frame with one row for each coefficient in the
original model. There are three columns, one for each of the tests:
Waldthe statistic is:
$$\frac{\hat{B}}{\hat{SE}}$$
where $\hat{B}$ is the estimate of the coefficient
and $\hat{SE}$ is its standard error.
plrPartial likelihood ratio test.
The statistic is the difference in the
likelihood ratio of the original model and that with the coefficient
omitted.
lrtThe log-rank test, aka the score test.
The Null hypothesis is that
$\hat{B}=0$.
The statistic is caculated by refitting the model with the coefficient
omitted, to generate initial values for the other $\hat{B}$s.
It is then fitted again with all
covariates, using these values and setting $\hat{B}=0$.
All statistics are distributed as $\chi$-square, with degrees of freedom
$=$ number of coefficients $-1$.