| Summary: | 碩士 === 國立嘉義大學 === 行銷與運籌研究所 === 100 === Gravity models have been widely applied to analyze bilateral trade between countries or areas. To calibrate a gravity model, (parametric) linear regression techniques have been commonly employed to develop the relationship between trade and economic factors. However, parametric regression models have their own model assumptions and pre-defined underlying relationship between dependent and independent variables. If these assumptions are violated, the model could lead to erroneous estimation of the trade between countries. Multivariate adaptive regression splines (MARS), one of widely applied data mining techniques, has been commonly applied in business administration, industry, and engineering. MARS does not require any pre-defined underlying relationship between dependent variable and independent variables and have been shown to be a powerful tool particularly in dealing with prediction problems. This study collected data of bilateral trade in 2008 and 2009 between APEC countries. A parametric linear regression and a nonparametric MARS model were developed to establish the empirical relationship between trade and multiple factors, including distance, population, GPD, logistic performance and many other economic variables. The estimation results of the linear regression model the MARS model are similar in general and show that the GDP, distance, logistic performance, common language and many other factors are the key determinants of trade between two countries. By comparing the prediction performance between linear regression and MASR models, this study demonstrates that nonparametric MARS models are good alternative methods for analyzing bilateral trade between countries in the world.
|
|---|
