This function checks and sets the fitting options for smoothSurvReg. Its arguments can be used instead of ... in a call to smoothSurvReg.
smoothSurvReg.control(est.c = TRUE, est.scale = TRUE,
maxiter = 200, firstiter = 0, rel.tolerance = 5e-5,
toler.chol = 1e-15, toler.eigen = 1e-3,
maxhalf = 10, debug = 0, info = TRUE, lambda.use = 1.0, sdspline = NULL,
difforder = 3, dist.range = c(-6, 6), by.knots = 0.3,
knots = NULL, nsplines = NULL, last.three = NULL)A list with the same elements as the input except dist.range and by.knots
is returned.
If TRUE the G-spline coefficients are estimated. Otherwise, they are fixed
to the values given by init.c parameter of smoothSurvReg.
If TRUE the scale parameter \(\sigma\) is estimated. Otherwise,
it is fixed to the value given by init.scale parameter
of smoothSurvReg.
Maximum number of Newton-Raphson iterations.
The index of the first iteration. This option comes from older versions of this function.
(Relative) tolerance to declare the convergence. In this version of the function, the convergence is declared if the relative difference between two consecutive values of the penalized log-likelihood are smaller than rel.tolerance.
Tolerance to declare Cholesky decomposition singular.
Tolerance to declare an eigen value of a matrix to be zero.
Maximum number of step-halving steps if updated estimate leads to a decrease of the objective function.
If non-zero print debugging information.
If TRUE information concerning the iteration process is printed during the computation to the standard output.
The value of the tuning (penalty) parameter \(\lambda\) used
in a current fit by the smoothSurvReg.fit function.
Value of this option is not interesting for the user. The parameter
lambda of the function smoothSurvReg is more important for
the user.
Standard deviation of the basis G-spline. If not given it is determined
as 2/3 times the maximal distance between the two knots. If est.c = TRUE
and sdspline >= 1 it is changed to 0.9 to be able to satisfy the constraints
imposed to the fitted error distribution.
The order of the finite difference used in the penalty term.
Approximate minimal and maximal knot. If not given by knots the knots
are determined as c(seq(0, dist.range[2], by = by.knots), seq(0, dist.range[1], by = -by.knots)).
The sequence of knots is sorted and multiple entries are removed.
The distance between the two knots used when building a vector of knots if these
are not given by knots. This option is ignored if nsplines is not NULL.
A vector of knots.
This option is ignored at this moment. It is used to give the number of G-splines
to the function smoothSurvReg.fit.
A vector of length 3 with indeces of reference knots. The 'a' coefficient of
the knot[last.three[1]] is then equal to zero, 'a' coefficients
with indeces last.three[2:3] are expressed as a function of remaining
'a' coefficients such that resulting error distribution has zero mean and unit
variance. If maxiter > 0 last.three is determined after
the convergence is reached. If maxiter == 0 last.three is used
to compute variance matrices.
Arnošt Komárek arnost.komarek@mff.cuni.cz