The researchers developed a new method to deal with a problem called multicollinearity in linear regression models. When the variables used in the model are too closely related, the usual method becomes less accurate. The new method they created helps to improve the accuracy of predictions in these situations.