The objective of this work was to set up and test a multiple linear regression model applied to principal components for representative coffee crop yield series for three places in Southern Minas Gerais, based on the model proposed by Stewart et al. (1976), with new variables, represented by agrometeorological elements, besides the soil water depletion for the four quarterly periods in agricultural cycle (July to June). Since the number of observations was lower than the amount number of variables, we resorted to principal component analysis to reduce the dimension of this set of variables. The multiple linear regression analysis was applied to the first three principal components. In agreement with the tests, the model presented relative errors of estimates with high discrepancies and a tendency to overestimate productivity for the three places. However, it was verified that the estimates for the model tended to present behavior similar to observed data.
yield; agrometeorological modeling; principal components; coffee bean
