Finding suitable statistical model for gross domestic product and agriculture related factors

Abstract

Gross domestic product (GDP) is a measure of a country’s economic performance. It is the market value of all officially recognized final goods and services produced within a country in a year, or over a given period of time. GDP per capita is often used as an indicator of a country's material standard of living. The economy of Sri Lanka is highly depending on agriculture, industry and services. The Objective of this research study is to find a suitable statistical model for GDP and agricultural related factors such as: Tea, Rubber, Coconut, Paddy, Fishing and Others. For this purpose annual data were collected form the central bank report from 1974 to 2014. Multiple regression models such as linear model, log linear model and first difference of log linear models are analyzed using the statistical tests: normality test, unit root test, residual test, autocorrelation and serial correlation test, heteroscedasticity, multicollinearity test and stability test. Among these three models the best model is selected based on the statistical significance of the models. The first difference of log linear model satisfied almost all the statistical tests. Hence, the first difference of log linear model is more appropriate for GDP and agriculture related factors.

Keywords