An iterative plug-in algorithm for P-Spline regression
Sebastian Letmathe () and
Yuanhua Feng ()
Additional contact information
Sebastian Letmathe: Paderborn University
Yuanhua Feng: Paderborn University
No 151, Working Papers CIE from Paderborn University, CIE Center for International Economics
This paper proposes a new IPI- (iterative plug-in) rule for optimal smoothing for penalised splines with truncated polynomials. The IPI is based on a closed-form approximation to the optimal smoothing parameter. In contrast to a DPI- (direct plug-in) approach the current algorithm is fully automatic and self-contained. Our proposal is a fixpoint-search procedure and the resulting smoothing parameter is (theoretically) independent of the initial value. Like the DPI, the IPI-rule can be employed as a refining stage in order to improve the quality of other selection methods, e.g. Mallowâ€™s Cp, Cross Validation or Residual Maximum Likelihood. Some numerical features of P-Splines as well as the performance of the IPI-algorithm are examined in detail through a simulation study. Our results reveal that our proposal works very well. Practical relevance of the IPI is illustrated by different data examples.
Keywords: P-Splines; smoothing parameter; iterative plug-in; simulation (search for similar items in EconPapers)
JEL-codes: C14 C51 (search for similar items in EconPapers)
New Economics Papers: this item is included in nep-ecm
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (1) Track citations by RSS feed
Downloads: (external link)
This item may be available elsewhere in EconPapers: Search for items with the same title.
Export reference: BibTeX
RIS (EndNote, ProCite, RefMan)
Persistent link: https://EconPapers.repec.org/RePEc:pdn:ciepap:151
Access Statistics for this paper
More papers in Working Papers CIE from Paderborn University, CIE Center for International Economics Contact information at EDIRC.
Bibliographic data for series maintained by WP-WiWi-Info () and ().