Calculates minus the log likelihood function and its first and second order derivatives for data from a Weibull regression model.
phfunc( beta = NULL, lambda, p, X = NULL, Y, offset = rep(0, length(Y)), ord = 2, pfixed = FALSE, dist = "weibull" )
| beta | Regression parameters |
|---|---|
| lambda | The scale paramater |
| p | The shape parameter |
| X | The design (covariate) matrix. |
| Y | The response, a survival object. |
| offset | Offset. |
| ord | ord = 0 means only loglihood, 1 means score vector as well, 2 loglihood, score and hessian. |
| pfixed | Logical, if TRUE the shape parameter is regarded as a known constant in the calculations, meaning that it is not cosidered in the partial derivatives. |
| dist | Which distribtion? The default is "weibull", with the alternatives "loglogistic" and "lognormal". |
A list with components
The log likelihood. Present if
ord >= 0
The score vector. Present if ord >= 1
The negative of the hessian. Present if ord >= 2
Note that the function returns log likelihood, score vector and minus hessian, i.e. the observed information. The model is $$S(t; p, \lambda, \beta, z) = S_0((t / \lambda)^p)^{e^(z \beta)}$$
Göran Broström