Auxiliary function for setting storage dimensions and other parameters
for rq.fit.sfn(), rq.fit.sfnc() and rqss().
sfn.control(nsubmax = NULL, tmpmax = NULL, nnzlmax = NULL, cachsz = 64,
small = 1e-06, maxiter = 100,
tiny = 1e-30, Large = 1e128,
warn.mesg = TRUE)A list with components named as the arguments given above.
upper bound for dimension of lindx
upper bound for dimension of tmpvec
upper bound for non-zero entries of L stored in lnz, including diagonal
size of cache in kbytes on target machine
convergence tolerance for interior point algorithm
maximal number of interior point iterations
a tiny positive number; values below tiny * max(diag) are replaced by Large;
originally was \(10^{-30}\) hardcoded in Fortran code.
a large number, practically “Infinite” to replace
tiny diagonal entries in Cholesky; was \(10^{128}\), hardcoded in
compiled code.
logical flag controlling printing of warnings.
Roger Koenker
Sparse fitting requires a number of temporary storage arrays whose size depends on problem specific features in somewhat mysterious ways, parameters controlling these sizes and some other fitting aspects can be controlled by specifying elements of this control object.
rq.fit.sfn,
rq.fit.sfnc,
and rqss from which sfn.control() is called.