Additional control options for the elastic net Peña-Yohai procedure.
enpy_options( max_it = 10, keep_psc_proportion = 0.5, en_algorithm_opts, keep_residuals_measure = c("threshold", "proportion"), keep_residuals_proportion = 0.5, keep_residuals_threshold = 2, retain_best_factor = 2, retain_max = 500 )
maximum number of EN-PY iterations.
how many observations should to keep based on the Principal Sensitivity Components.
options for the LS-EN algorithm. See en_algorithm_options for details.
how to determine what observations to keep, based on their residuals.
proportion of observations to kept based on their residuals.
only observations with (standardized) residuals less than this threshold are kept.
only keep candidates that are within this factor of the best candidate. If
maximum number of candidates, i.e., only the best
options for the ENPY algorithm.
The EN-PY procedure for computing initial estimates iteratively cleans the data of observations with possibly outlying residual or high leverage. Least-squares elastic net (LS-EN) estimates are computed on the possibly clean subsets. At each iteration, the Principal Sensitivity Components are computed to remove observations with potentially high leverage. Among all the LS-EN estimates, the estimate with smallest M-scale of the residuals is selected. Observations with largest residual for the selected estimate are removed and the next iteration is started.