In a classical multiple linear regression analysis, multicollinearity and autocorrelation are two main basic assumption violation problems. When multicollinearity exists, biased estimation techniques such as Maximum Likelihood, Restricted Maximum Likelihood and most recent the K-L estimator by Kibria and Lukman [1] are preferable to Ordinary Least Square. On the other hand, when autocorrelation exist in the data, robust estimators like Cochran Orcutt and Prais-Winsten [2] estimators are preferred. To handle these two problems jointly, the study combines the K-L with the Prais-Winsten’s two-stage estimator producing the Two-Stage K-L estimator proposed by Zubair & Adenomon [3]. The Mean Square Error (MSE) and Root Mean Square Error (RMSE) criterion was used to compare the performance of the estimators. Application of the estimators to two (2) real life data set with multicollinearity and autocorrelation problems reveals that the Two Stage K-L estimator is generally the most efficient.