Caribbean food imports often face detentions and refusals by the US resulting in a major loss of income. In this paper, we consider a classification procedure based on transvariation probabilities to correctly identify cases that lead to food detention. This is based upon several background variables on fourteen Latin American and Caribbean countries. A method for selecting variables according to their contribution towards predicting detention is given. For our particular sample, the selection method chose foreign direct investment as the variable that carried the most information in determining food detention. After removing variables that were non-informative about food detention status, a leave-one-out cross-validation shows that methods based on transvariation probabilities were superior to classical methods in predicting food detention.