To develop a multivariate model for predicting yield data were collected in farmers’ sorghum fields where gray leaf spot, smut, rust, and oval leaf spot epidemics occurred in 1985. Several models were tested to explain variation in yield based on a range of variables measured on individual plants. Data were first subjected to principal component analysis, and the major principal components were used as independent variables in a multiple regression model, with yield the dependent variable. The best model included the following variables: per cent severity of the 4 diseases, panicle length, and plant height. Using the model, yield differences of up to 7% were predicted when comparing the minimum‐ and the maximum‐disease severities observed and holding plant height and panicle length at their mean values. Using the observed plant height and panicle length corresponding to actual disease severities, yield differences estimated were 14.6%, 3.6%, and 5.5% for gray leaf spot, rust, and oval leaf spot, respectively. Actual losses from smut were insignificant.