pencoxfrailControl: Control Values for 'pencoxfrail' fit

In PenCoxFrail: Regularization in Cox Frailty Models

Control Values for pencoxfrail fit

Description

The values supplied in the function call replace the defaults and a list with all possible arguments is returned. The returned list is used as the control argument to the pencoxfrail function.

Usage

pencoxfrailControl(start = NULL, q_start = NULL, conv.eps = 1e-4, 
                          standardize = FALSE, center = FALSE,
                          smooth=list(nbasis = 6, penal = 0.1), 
                          ridge.pen = 1e-4, print.iter = FALSE, 
                          max.iter = 100, c.app = 1e-6, zeta = 0.5, 
                          exact = 1e-2, xr = NULL, ...)

Arguments

start	a vector of suitable length containing starting values for the spline-coefficients of the baseline hazard and the time-varying effects, followed by the fixed and random effects. The correct ordering is important. Default is a vector full of zeros.
q_start	a scalar or matrix of suitable dimension, specifying starting values for the random-effects variance-covariance matrix. Default is a scalar 0.1 or diagonal matrix with 0.1 in the diagonal, depending on the dimension of the random effects.
conv.eps	controls the speed of convergence. Default is 1e-4.
center	logical. If true, the covariates corresponding to the time-varying effects will be centered. Default is FALSE (and centering is only recommended if really necessary; it can also have a strong effect on the baseline hazard, in particular, if a strong penalty is selected).
standardize	logical. If true, the the covariates corresponding to the time-varying effects will be scaled to a variance equal to one (*after* possible centering). Default is FALSE.
smooth	a list specifying the number of basis functions nbasis (used for the baseline hazard and all time-varying effects) and the smoothness penalty parameter penal, which is only applied to the baseline hazard. All time-varying effects are penalized by the specific double-penalty \xi\cdot J(\zeta,\alpha) (see pencoxfrail), which is based on the overall penalty parameter \xi (specified in the main function pencoxfrail) and on the weighting between the two penalty parts \zeta. The degree of the B-splines is fixed to be three (i.e. cubic splines).
ridge.pen	On all time-varying effects (except for the baseline hazard) a slight ridge penalty is applied on the second order differences of the corresponding spline coefficients to stabilize estimation. Default is 1e-4.
print.iter	logical. Should the number of iterations be printed? Default is FALSE.
max.iter	the number of iterations for the final Fisher scoring re-estimation procedure. Default is 200.
c.app	The parameter controlling the exactness of the quadratic approximations of the penalties. Default is 1e-6.
zeta	The parameter controlling the weighting between the two penalty parts in the specific double-penalty \xi\cdot J(\zeta,\alpha) (see pencoxfrail). Default is 0.5.
exact	controls the exactness of the (Riemann) integral approximations. Default is 1e-2.
xr	maximal time point that is regarded. Default is NULL and the maximal event or censoring time point in the data is used.
...	Futher arguments to be passed.

Value

a list with components for each of the possible arguments.

Author(s)

Andreas Groll [e-mail]

See Also

pencoxfrail

Examples

# Use different weighting of the two penalty parts
# and lighten the convergence criterion 
pencoxfrailControl(zeta=0.3, conv.eps=1e-3)

PenCoxFrail documentation built on Sept. 11, 2024, 7:12 p.m.