Wang Guoxin - Engineer,Seismological Bureau, Liaoning Province, China
Guo Yuxue - Professor, Institute of Engineering Mechanics,Herbin,China

This paper presents a improved method--Robust Regression(RR) to deal with the problems existing larger differences between results and most observed data, especially when some outliers are included in the data, and discusses the possibility and necessity of using RR in earthquake engineering. THE NECESSARY OF USING RR IN EARTHQUAKE ENGINEERING The Least Squares Method(LSM) is commonly used in many aspects,because it has some advantages, such as surtability,and most results obtained from LSM can reflect the real situation.However,peoples can also find the differences between results and most observed data, and try to improve these by other ways. For example, to determine the player's score In the compeation of sports or singers, the highest and the lowest ones(commonly one or two) are often omitted among the total scores, and then to take the mean value of leaving ones or median as the final score.These methods are the improvement to the statistical one,and also raise the reliability and robustness.