|
| << vorige | volgende >> |
Tijdschrift beschrijving
Alle jaargangen van het bijbehorende tijdschrift
Alle afleveringen van het bijbehorende jaargang
Alle artikelen van de bijbehorende aflevering
Details van artikel 4 van 20 gevonden artikelen
| Titel: |
Binned modified cross-validation with dependent errors |
| Auteur: |
Chu, C.K. |
| Verschenen in: |
Communications in statistics |
| Paginering: | Jaargang 23 (1994) nr. 12 pagina's 3515-3537 |
| Jaar: |
1994 |
| Inhoud: |
For nonparametric regression, in the case of dependent observations, the performance of both the kernel estimator and its associated bandwidth selector applied to binned data is investigated. We show that binning the data has no effect on the asymptotic mean squared error of the kernel estimator. But it tends to alleviate the effect of dependence on cross-validation. The amount of the dependence effect decreases as the bin size increases. The performance of one method, binned modified cross-validation (BMCV), which adjusts for the effect of dependence on bandwidth selection is also studied. The limiting distribution for the bandwidth produced by BMCV is given. In the case of positively correlated data, BMCV has stronger effects of reducing both bias and variability in the selected bandwidth than modified cross-validation, when each method leaves out the same number of observations. This result still holds in other cases if the number of observations left out is sufficiently large. Simulations demonstrate that the asymptotic effects hold for reasonable sample sizes. |
| Uitgever: |
Taylor & Francis |
| Bronbestand: |
Elektronische Wetenschappelijke Tijdschriften |
Details van artikel 4 van 20 gevonden artikelen
| << vorige | volgende >> |