Statistical Modeling for Area of Pearl Millet in Gujarat, India
Jignesh K. Parmar *
Department of Agricultural Statistics, B.A.C.A., AAU, Anand – 388110, India.
A. D. Kalola
Department of Agricultural Statistics, B.A.C.A., AAU, Anand – 388110, India.
Darshan L. Kothiya
Department of Agricultural Statistics, B.A.C.A., AAU, Anand – 388110, India.
Krunal Panchal
University of Massachusetts, Boston, USA.
*Author to whom correspondence should be addressed.
Abstract
The study focuses on modeling the area under pearl millet cultivation in Gujarat, India, using various statistical methods. The analysis employs both parametric (ARIMA, linear, non-linear and Exponential Smoothing State Space models) and non-parametric (LOESS smoothing) approaches. Data from 1980-81 to 2021-22 were used, with models trained on data up to 2013-14 and tested on subsequent data. The study found that the cubic and logistic models provided the best fit among linear and non-linear models, respectively, while ARIMA (0,1,1) with drift and ETS (A,A,N) were optimal among time series models. Ultimately, the cubic model was slightly superior, though ARIMA (011) with drift showed the best forecasting ability. Despite efforts like the National Millet Mission and International Year of Millets, the area under pearl millet cultivation continues to decline, indicating a need for enhanced strategies to support pearl millet farming in Gujarat.
Keywords: Non-linear and linear models, auto-correlation, band width, kernel, pearl millet, statistical modeling, ARIMA, exponential smoothing, LOESS smoothing