Multiplicative local linear hazard estimation and best one-sided cross-validation
Authors: Maria Luz Gámiz, María Dolores Martínez-Miranda, Jens Perch Nielsen
JMLR 2019 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Best one-sided cross-validation turns out to have excellent performance in its practical illustrations, in its small sample performance and in its mathematical statistical theoretical performance. Two case studies show the applicability of our proposals, which are described in Section 6. In Section 7 we describe simulation experiments to evaluate the finite sample properties of our proposal. |
| Researcher Affiliation | Academia | Mar ıa Luz G amiz EMAIL Department of Statistics and Operations Research, University of Granada Campus Fuentenueva s/n, 18071 Granada, Spain; Mar ıa Dolores Mart ınez-Miranda EMAIL Department of Statistics and Operations Research, University of Granada Campus Fuentenueva s/n, 18071 Granada, Spain; Jens Perch Nielsen EMAIL Cass Business School, City, University of London 106 Bunhill Row, London EC1Y8TZ, U.K. |
| Pseudocode | No | The paper only describes methods and procedures using mathematical formulations and textual descriptions, without presenting any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | All numerical calculations have been performed with R and the methods proposed in this paper have been implemented in the DOvalidation package (G amiz et al., 2017). |
| Open Datasets | Yes | Our first application is on fitting hazard mortality curves for old-age population. We consider mortality data of women in Iceland in the calendar year 2006, with ages from 40 to 110. The same data were considered by G amiz et al. (2016) and are available in the DOvalidation R-package (G amiz et al., 2017). |
| Dataset Splits | Yes | The cross-validated bandwidth, denoted by bb CV,K, is therefore the minimizer of b QK(b) = n 1 " n X 0 {bαb,K(s)}2 Yi(s)w(s)ds 2 0 bα[i] b,K(s)w(s)d Ni(s) where bα[i] b,K(s) is the estimator arising when the data set is changed by setting the stochastic process Ni(s) equal to 0 for all s [0, T]. |
| Hardware Specification | No | The paper mentions computations were performed by Centro de Servicios de Inform atica y Redes de Comunicaciones (CSIRC), University of Granada, but does not specify any particular hardware details such as CPU/GPU models, memory, or other specifications used for the experiments. |
| Software Dependencies | Yes | All numerical calculations have been performed with R and the methods proposed in this paper have been implemented in the DOvalidation package (G amiz et al., 2017). |
| Experiment Setup | Yes | For models 1 to 4 we have considered sample sizes n = 100, 1000, 10000, and for model 5, n = 50000, 75000, 100000. The number of Monte Carlo replications for each case has been always 500. ... The grid size has been chosen equal to R = 500 in both cases. ... we have calculated the local linear hazard estimator and its multiplicative bias correction using the sextic kernel: K(x) = 3003/2048(1 x2)6I( 1 < x < 1)... To compute all bandwidth estimates we have considered grids of 100 equally spaced bandwidth values chosen around bb ISE, for each model and sample size. All criteria have been defined using a weighting function such that w(s)Y (s) 1... |