Auxiliary function for iCoxph that enable users
to specify the starting values of the model estimation procedure.
iCoxph.start(
beta_vec = NULL,
beta_mat = NULL,
methods = c("nearest_hazard", "unit_hazard"),
...
)A numeric vector for starting values of coefficient estimates. The default values are the coefficient estimates from the regular Cox model only fitting on records without uncertainty. If censoring rate among subjects having unique certain records is extremely high (> 99 indicator and one predictor, the starting values will be reset to be all zeros.
A numeric matrix that consists of additional starting values
of coefficient estimates in columns. The default value is NULL.
A character vector specifying the initialization methods for
probabilities of uncertain records being true. The available methods
are "nearest_hazard" for initializing baseline hazard by nearest
(left) neighbor, and "unit_hazard" for initializing unit baseline
hazard. Partial matching on method names is supported for ease of
typing. By default, both methods are used. See Wang et al. (2020) for
details of the initialization methods.
Other arguments for future usage. A warning will be thrown if any invalid argument is specified.
A list of class intsurv-iCoxph.start containing all specified
starting values of the parameters to be estimated from the model.
iCoxph for fitting integrative Cox model.
# NOT RUN {
## See examples of function 'iCoxph'.
# }
Run the code above in your browser using DataLab