IMPORTANCE OF MELON TYPE, SIZE, GRADE, CONTAINER, AND SEASON IN DETERMINING MELON PRICES

Classification and Regression Trees (CART), a computer intensive nonpa rametric classification method, was used to model weekly Los Angeles w holesale prices (1990-93) for twelve different melon types. CART expla ined more of the variation in melon prices than did an ordinary least squares (OLS) regression with dummy variables. Explanatory variables r anked as the most-to-least important by CART are as follows: week, typ e of melon, year, size, grade, and shipping container. The most notabl e price change occurs when prices fall after 13 May.