The study compared Least Squares Regression and Principal Components Regression when dealing with multicollinearity in datasets. They created 10 datasets with different levels of multicollinearity and found that in Least Squares Regression, the sign of some coefficients was reversed due to multicollinearity, while in Principal Components Regression, the coefficients were in the right direction. The standard errors of the coefficients in Principal Components Regression were also lower than in Least Squares Regression. Therefore, it is recommended to use Principal Components Regression instead of Least Squares Regression when multicollinearity is present in the data.