Advances and Applications in Statistics
Volume 55, Issue 2, Pages 193 - 220
(April 2019) http://dx.doi.org/10.17654/AS055020193 |
|
PARSIMONIOUS MODEL FOR PREDICTING COCOA PRODUCTION IN GHANA
A. Buabeng, S. Twumasi-Ankrah, K. A. Nyantakyi and B. Kumi-Boateng
|
Abstract: This study examined the combined effect of socio-economic and climatic variables on Ghana’s cocoa production. First, only multiple regression analysis was applied to all the variables affecting cocoa production which caused multicollinearity problems. In order to eliminate multicollinearity problems and perform a reliable regression, factor analysis was employed after its appropriateness on the data has been tested. Thus the problems were removed by using factor scores, summated scales and surrogate variables for the regression analysis. Also, the most significant determinants affecting cocoa production in Ghana was identified through which various parsimonious models were developed using regression models with ARIMA errors technique. The model parameters had least multicollinearity values that best describe and predict Ghana’s cocoa production. The parsimonious models were then compared in terms of prediction accuracy. The factor scores model (with an interaction term) was concluded to give better interpretation and good estimates of Ghana’s cocoa production as compared to the remaining models. |
Keywords and phrases: multicollinearity, factor analysis, factor scores, summated scales, surrogate variables, regression models with ARIMA errors technique, parsimonious.
|
|
Number of Downloads: 313 | Number of Views: 1203 |
|