Predicting adoption of double cropping in paddy fields of northern Iran: a comparison of statistical methods

Factors affecting the adoption of double cropping were explored in rice farms of Fouman County of Guilan Province in northern Iran using artificial neural networks (ANNs), linear discriminant analysis (LDA), and logistic regression (LGR). Eleven factors (age, education, occupation, family size, type of farm ownership, distance to the agricultural service center, attending agricultural extension courses, use of financial resources and bank loans, number of domestic animals, area under cultivation, and social participation) were examined. An additional objective was to compare the ability of the three models in predicting the adoption of double cropping. ANNs showed an overall predictive power of 89.8%. LDA showed an overall predictive power of 83.2%, with seven of the eleven independent variables being effective on the adoption of double cropping. LGR indicated an overall predictive power of 87.6%, with eight of the eleven independent variables being effective on the adoption of double-rice cropping. ANNs showed higher power than LGR and LDA in predicting the adoption of double cropping. Based on all three methods used for analysis, the most important independent variables were social participation and area under cultivation (positive factors) as well as distance to the agricultural service center and family members (negative factors). Establishment of cooperatives or other kinds of farmers' associations to foster social participation could motivate adoption of double cropping, particularly among small-scale farmers. To increase agricultural services, more local centers should be created in rural areas. The government should promote double cropping through effective incentives and technology transfer to small-scale farmers.