Statistics for Agricultural credit models: identifying high risk applications