USING AN ARIMA MODEL TO FORECAST BOVINE MILK PRODUCTION IN BAJA CALIFORNIA, MEXICO

Abstract

An ARIMA model was constructed to represent the behavior of local milk production, using the monthly data between January of 2000 and December of 2009 for milk production in the state of Baja California, Mexico. The three step methodology described by Box and Jenkins for model selection and estimation was applied. Both, the correlogram behavior and the Dickey-Fuller test were used to determine if the series was stationary. The series resulted stationary and no differencing needed, then the process of model selection was performed and as result of this two ARMA models (1,1 and 2,2) were proposed. The parameters of model were estimated by ordinary least squares and the statistics and regression results in conjunction with the Akaike and Schwarz criteria (18.06, 18.13 y 18.20, 18.27, respectively) with the results of the correlogram of each model were used to determine which model better represented the data generating process. The results indicated the (1,1) model was better in fitting, thus this model had better forecasting characteristics. It was concluded that this type of time series model can be considered as a useful tool to describe and forecast milk production however it usefulness is restricted to short term forecasts.