PROBLEM OF AUTOCORRELATION IN LINEAR REGRESSION DETECTION AND REMEDIES

In the classical linear regression model we assume that successive values of the disturbance term are temporarily independent when observations are taken over time. But when this assumption is violated then the problem is known as Autocorrelation. When the disturbance term exhibits serial correlation, the value of the standard error of the parameter estimates are affected and the predictions based on ordinary least square estimates will be inefficient. So we cannot estimate the correct values of the parameters and the estimates are biased. In this study main focus is given on how one can detect the problem of autocorrelation and how the problem can be solved so as to we can estimate the values of the parameters correctly that are best, linear and unbiased. To explain the procedure of detection of autocorrelation and its remedial measure we take an example on indices of real compensation per hour and output per hour in the business sector of the U.S economy for the period 1960-2005. We use Run Test and Durbin-Watson test to detect the problem of autocorrelation, and then explain the procedure how the problem can be solved.