COMPARISON OF TWO REGRESSION MODELS FOR PREDICTING CROP YIELD

The linear regression model based on the ordinary least square (OLS) estimation is a commonly used method for crop yield predicting. But it is not adequate in many cases because spatial autocorrelation among variables may violate the underlying assumption that observations are independent. In this study, we compared the OLS regression model and the spatial autoregressive model for predicting corn yield in Iowa. The spatial autoregressive model indicated a significant improvement in model performance over OLS model. The spatial autoregressive model can provide better prediction than OLS model and is capable of adjusting the spatial autocorrelation, which is often ignored by the OLS model. The study demonstrated that NDVI and precipitation are the major predictors for forecasting corn yield in Iowa.