An object returned by the curelps function consists in a list
with various components related to the fit of a promotion time cure model
using the Laplace-P-spline methodology.
A curelps object has the following elements:
The formula of the promotion time cure model.
Number of B-spline basis functions used for the fit.
Chosen penalty order.
The dimension of the latent field. This is equal to the sum of the number of B-spline coefficients and the number of regression parameters related to the covariates.
The observed event times.
Sample size.
The number of events that occurred.
The upper bound of the follow up, i.e. max(event.times).
The event indicators.
Posterior estimates of the regression coefficients related to the cure probability (or long-term survival).
Posterior estimates of the regression coefficients related to the population hazard dynamics (or short-term survival).
The maximum a posteriori of the (log-)posterior penalty parameter.
The quadrature points of (log-) posterior penalty parameters used to compute the Gaussian mixture posterior of the latent field vector.
The estimated B-spline coefficients.
Estimated effective degrees of freedom for each latent field variable.
The effective model dimension.
The posterior covariance matrix of the B-spline coefficients for a penalty fixed at its maximum posterior value.
The posterior covariance matrix of the latent field for a penalty fixed at its maximum posterior value.
The covariate matrix for the long-term survival part.
The covariate matrix for the short-term survival part.
The log-likelihood evaluated at the posterior latent field estimate.
Number of parametric coefficients in the model.
The AIC computed with the formula -2*loglik+2*p, where p is the number of parametric coefficients.
The AIC computed with the formula -2*loglik+2*ED, where ED is the effective model dimension.
The BIC computed with the formula -2*loglik+p*log(ne), where p is the number of parametric coefficients and ne the number of events.
The BIC computed with the formula -2*loglik+ED*log(ne), where ED is the effective model dimension and ne the number of events.
Oswaldo Gressani oswaldo_gressani@hotmail.fr.
curelps