Herawati, Netti and Nisa, Khoirin and Setiawan, Eri (2017) THE OPTIMAL BANDWIDTH FOR KERNEL DENSITY ESTIMATION OF SKEWED DISTRIBUTION: A CASE STUDY ON SURVIVAL TIME DATA OF CANCER PATIENTS. Prosiding Seminar Nasional METODE KUANTITATIF 2017. ISSN 978-602-98559-3-7

[img]
Preview
Text
MK NHKNES.pdf

Download (1MB) | Preview

Abstract

In this paper, optimal bandwidth selection for skewed distribution is studied through data simulation. The data are generated from Exponential (1), Exponential (5), Gamma (1,6), Gamma (1,9), Weibull (1,5), and Weibull (1,10) distributions having parameter(s) that produces a skewed density function with n = 100. The Gaussian kernel density functions of the generated distributions are estimated using Scott (Nrd), Silverman’s rule of thumb (Nrd0), Silverman’s Long-Tailed distribution (Silverman-LT), Biased Cross validation (BCV) and Sheater-Jones (SJ) bandwidth methods. The kernel density estimates are compared to the corresponding probability density functions. The selected optimal bandwidth then is applied to kernel density estimation of survival time data of cancer patients. Result indicates that, overall, Silverman’s Rule of Thumb (Nrd0) method outperformed the other methods.

Item Type: Article
Subjects: H Social Sciences > HA Statistics
Depositing User: DR. KHOIRIN NISA
Date Deposited: 27 Mar 2018 04:46
Last Modified: 27 Mar 2018 04:46
URI: http://repository.lppm.unila.ac.id/id/eprint/6650

Actions (login required)

View Item View Item