Comparative Analysis of Regression Models and Methods of Estimating Unbiased Parameters for Crop Production

Abstract

This work was to establish a predictive equation for crop production. It also provide empirical information and examination on comparative analysis of regression models and methods of estimating unbiased parameters (i.e. six methods) with focus on crop production. Time series secondary data extracted from a survey data carried out in Nigeria on crop production from the National Bureau of Statistics from January 2010 to December 2021 were employed. The analyses of the data were done using these six methods: Arithmetic Mean, Geometric Mean, Harmonic Mean, Wald, Bartlett, and Durbin's Methods. All computations were done via Microsoft Excel, Gretl (version 18.0) and MINITAB (version 20.0). The general regression model is given by Yi – β0 + β1x1– β2x2 + ɛi (cassava and rice production are the independent variables; "x1 " jand "x2 " while crop production is the dependent variable "yi"). Then, the actual model, centre of mean models and instrumental variables methods models were estimated. The result of the parameters estimated for all the methods models are significant at 5%. In addition, the models' residuals sum of square (RSS), R-square value (R2), Akaike Information criteria (AIC) and Bayesian Information Criterion (BIC) of the models were calculated and compared. The results of the comparison identify the Model F as the best model which parameter estimates was obtained by the Bartlett Method. Hence, Bartlett Method of estimators is the best-unbiased parameters estimation method using the model selection criteria considered that is R-square value (R2), AIC and BIC. This paper recommended that the six methods of estimating unbiased parameters should be used with other crop produce to identify the model with the estimators of the best-unbiased parameters.