ESTIMATION OF TRADE BALANCE USING MULTIPLE LINEAR REGRESSION MODEL

Abstract

This study aims to evaluate the performance of multiple linear regression in estimating trade balance, so that a regression model for estimating the trade balance can be developed based on the important variables that have been identified. The performance of four regression methods including enter, stepwise regression, backward deletion, and forward selection is measured by mean absolute error, standard deviation, and Pearson correlation at the validation stage. The study concludes that multiple linear regression model developed by stepwise method is the best model for the trade balance estimation. The model considers the following six significant variables: Exports of palm oil, imports of tubes, pipes, and fittings of iron or steel, exports of crude petroleum, imports of petroleum products, exports of plywood plain, and imports of rice. The regression model achieves a moderate value of model estimated accuracy (76.10%), mean absolute error (0.257), standard deviation (0.308), and linear correlation (0.851).