Kesesuaian Metode Regresi Nonparametrik Spline, B-spline , dan P-spline dalam Menduga Kurva Regresi

Sugiarti, Harmi (2012) Kesesuaian Metode Regresi Nonparametrik Spline, B-spline , dan P-spline dalam Menduga Kurva Regresi. Project Report. Universitas Terbuka, Jakarta.

[img]
Preview
Text
2012_97.pdf

Download (2MB) | Preview

Abstract

Regression modeling with nonparametric regression method is required as an alternative method of standard methods (parametric regression method) to obtain a more accurate model when there is assumption “true model” violation. There are several nonparametric regression methods that can be used, including the spline methods. This study aims to assess performance of the spline basis method with truncated (spline method) , spline method with B-spline basis (B-spline), and penalized spline method with B-spline basis (P-spline). This study used simulated data and applied data. Simulated data are generated and processed with the R programs version 2.14.1. Applied data are obtained from the SRS-UT with the first grade point and life time study for Statistics Department students in 1989.2-2011.2 periods. Base on the 100 paired observations from generated data and the generalized cross validation (GCV) and mean squared error (MSE) criterion , the spline method and B-spline method with three knots have good performance to fit the data, while the P-spline method have optimal performance when 10 knots are used. Base on 835 paired observations from real data, there is not significantly different performance of the spline method, B-spline method, and the P-spline method.

Item Type: Monograph (Project Report)
Additional Information (ID): 2012_97
Uncontrolled Keywords: metode regresi nonparametik spline, b-apline, p-spline, kurva regresi
Subjects: 500 Natural Science and Mathematics > 510-519 Mathematics (Matematika) > 519.53 Descriptive Statistical Mathematics/Matematika Statistikal Deskriptif
Divisions: Thesis,Disertasi & Penelitian > Penelitian
Depositing User: Praba UT
Date Deposited: 08 Nov 2016 02:01
Last Modified: 08 Nov 2016 02:01
URI: http://repository.ut.ac.id/id/eprint/5774

Actions (login required)

View Item View Item