The researchers developed a method to improve the accuracy of predicting variables by using partial least squares regression. They identified and removed outliers in the data, which are points that greatly deviate from the norm, to create a more reliable model. By doing this, they found that the precision of the model significantly increased.