Auxiliary function for controlling Qcoxph fitting. Estimation proceeds in three steps: (i) evaluation of starting points; (iia) stochastic gradient-based optimization (iib) standard gradient-based optimization; and (iii) Newton-Raphson. Step (i) is based on a preliminary fit of a Cox model via coxph. Steps (iia) and (iib) find an approximate solution, and make sure that the Jacobian matrix is well-defined. Finally, step (iii) finds a more precise solution.
Qcoxph.control(tol = 1e-8, maxit, safeit, alpha0, display = FALSE)A list with named elements as in the argument list
tolerance for convergence of Newton-Raphson algorithm, default is 1e-8.
maximum number of iterations of Newton-Raphson algorithm. If not provided, a default is computed as 50 + 25*npar, where npar is the total number of parameters.
maximum number of iterations of gradient-search algorithm. If not provided, a default is computed as 10 + 5*npar, where npar is the total number of parameters.
step size for the preliminary gradient-based iterations. If estimation fails, you can try choosing a small value of alpha0. If alpha0 is missing, an adaptive choiche will be made internally.
Logical. If TRUE, tracing information on the progress of the optimization is printed on screen. Default is FALSE.
Gianluca Sottile <gianluca.sottile@unipa.it> Paolo Frumento <paolo.frumento@unipi.it>
If called with no arguments, Qcoxph.control() returns a list with the current settings of these parameters. Any arguments included in the call sets those parameters to the new values, and then silently returns.
Qcoxph