Specify additional options for the initial estimator based on the Pena-Yohai estimator.
initest_options(keep_solutions = 5, psc_method = c("exact", "rr"),
maxit = 10, maxit_pense_refinement = 5, eps = 1e-06, psc_keep = 0.5,
resid_keep_method = c("proportion", "threshold"), resid_keep_prop = 0.6,
resid_keep_thresh = 2, mscale_eps = 1e-08, mscale_maxit = 200)how many initial estimates should be kept to perform full PENSE iterations?
The method to use for computing the principal sensitivity components. See details for the possible choices.
maximum number of refinement iterations.
maximum number of PENSE iterations to refine initial estimator.
numeric tolerance for convergence.
proportion of observations to keep based on the PSC scores.
How to clean the data based on large residuals.
If "proportion", observations with the smallest
resid_keep_prop residuals will be retained.
If "threshold", all observations with scaled residuals smaller
than the threshold resid_keep_thresh will be retained.
proportion or threshold for observations to keep based on their residual.
maximum number of iterations and numeric tolerance for the M-scale.
a checked options list.
Two different methods to calculate the sensitivity components are implemented:
"rr"Approximate the PSCs by using the residuals from the
elastic net fit and the hat matrix from the ridge regression.
This method only works if alpha < 1 or
ncol(x) < nrow(x).
"exact"Calculate the PSCs from the difference between the residuals and leave-one-out residuals from elastic net.
Pena, D., & Yohai, V.. (1999). A Fast Procedure for Outlier Diagnostics in Large Regression Problems. Journal of the American Statistical Association, 94(446), 434-445. http://doi.org/10.2307/2670164
Other specifying additional options: en_options_aug_lars,
mstep_options, pense_options