| Title | Local linear extrapolation |
| Publication Type | Journal Article |
| Year of Publication | 2003 |
| Authors | Li, XC, Heckman, NE |
| Journal | JOURNAL OF NONPARAMETRIC STATISTICS |
| Volume | 15 |
| Pagination | 565-578 |
| Date Published | AUG-OCT |
| Type of Article | Article |
| ISSN | 1048-5252 |
| Keywords | bandwidth selection, local linear estimation, prediction |
| Abstract | A local linear estimator is proposed for extrapolating beyond the range of independent regression data. The asymptotic bias and variance of the local linear extrapolant are presented and used to calculate the asymptotic mean squared error. Three procedures are considered for choice of the bandwidth: a plug-in procedure. a forward cross-validation procedure and a procedure that combines the two. All of these are, in some sense. based on minimizing the asymptotic mean squared error. A simulation study shows that the forward cross-validation procedure outperforms the other two procedures both in terms of choosing a bandwidth and in terms of accurate extrapolation. The techniques are applied to two data sets. |
| DOI | 10.1080/10485250310001605432 |
