Estimation of Multivariate Adaptive Regression Splines (MARS) Model Parameters by Using Generalized Least Square (GLS) Method

Nurul Azizah Rahmadani Ritonga

Abstract


The regression analysis method for estimating the regression curve is divided into 3 (three) categories, namely parametric regression analysis, non-parametric regression analysis, and semi-parametric regression analysis. One form of non-parametric regression model is spline which can be developed into Multivariate Adaptive Regression Splines (MARS). The OLS estimation method will get good estimation results compared to other methods if the classical assumptions are fully met. However, if the classical assumptions cannot be fulfilled, this method is not good enough to use. The GLS method can be used if the classical assumptions required by the OLS method are not met. This study aims to estimate the parameters of the MARS model using the GLS method. The GLS method can be used if the classical assumptions required by the OLS method are not met. An example of a case used in the application of non-parametric estimation of the MARS model is the data on the number of doctors and gross enrollment rates for tertiary institutions in 32 districts/cities in North Sumatra in 2021. The best MARS model obtained in this study was obtained with a knot point of 21.2, 24 .2 and 27.2, with BF=6, MO=3, MI=0 with a GCV value of 6628.965. The best model obtained based on this research is as follows:

 


Keywords


Parameters Estimation; MARS Model; GLS Method

Full Text:

PDF

References


Astuti, Desak Ayu Wiri, 2016.Multivariate Spline Nonparametric Regression Analysis for Modeling Poverty Indicators in Indonesia. Mathematics E-Journal,Vol.5 (3), pp.111-116. ISSN:2303-1751

Breiman, L., Friedman, J., Olshen, R., & Stone, J. (1993). Classification and Regression Trees. New York: Chapman and Hall.

Budiantara, I. N. 2009. "U, GML, CV, and GCV Methods in Spline Nonparametric Regression",Scientific Magazine of the Indonesian Mathematical Association (MIHMI), vol. 6, hal.285-290.

Friedman, J.H,.1991. Multivariate Adaptive Regression Splines, The Annals of Statistics. 1991;19(1). Matter. 1-141.

Greene, W. H. 1997. Econometric analysis. Pearson Education, New York.

Gujarati, D. N. and Porter, D. C. 2009. Basic Econometrics. Fifth Edition. McGraw Hill/Irwin. New York,

Harini, S. and Turmudi. 2008.Method Statistics.Malang: UIN-Malang Press.

Hasan, M. I. 2002. The Main Materials of Statistics I (Descriptive Statistics).Jaka PT. Script Earth.

Tehupuring, B. C., & Hadi, S. (2015). Application of the Multivariate Adaptive Regression Spline as a Tool for Modeling the Growth of Broiler Chickens.VETERINARY Act Indonesiana, 3(1), 23-28.

Wasilane, T. 2014.Ridge Regression Model To Overcome Multiple Linear Regression Models Containing Multicollinearity. Barekeng Journal, 31-37.

Zhang, H., & Singer, B. H. 2010. Recursive partitioning and applications. Springer Science & Business Media.

Zurimi, S. 2019. Analysis of the Multivariate Adaptive Regression Spline (MARS) applicative model on the classification of factors that influence the study period of FKIP students Ambon Darussalam University.Symmetric Journal, 9(2), 250-255.




DOI: https://doi.org/10.30596/jmea.v2i2.13106

Refbacks

  • There are currently no refbacks.


 

 

Journal of Mathematics Education and Application: JMEA

University Muhammadiyah of Sumatera Utara

Magister Pendidikan Matematika Program Pascasarjana Universitas Muhammadiyah Sumatera Utara, Jl. Denai No 217, Indonesia

email: jmea@umsu.ac.id