An alternative approach to least squares estimators for time dependent AR(p) errors, using regression quantiles

The in dependence of measurements for response variables may not exist when the variables are likely to drift over time such as the results of a medical test under some continuous treatments or stock prices of economic indicators which depend on time durations. In these situations the errors may follow the AR (p) model. We try to find an alternative for least squares method. AR (1) case has been treated by Ullah and Srivastava etc. (1983) using two stage and J.W. Galbraith (1991) carried out a consistent estimator of linear regression parameters using maximum likelihood method. One of the approaches for solving the problem was based on robust estimators such as regression quantiles. We use the least absolute diviation estimator (LADE), which is a robust methods. To obtain an alternative approach for estimating regression pararameters under AR(p) time dependent errors.

کلیدواژه ها:

Regression Quantiles ، LADE ، LSE ، AR(p)