EN
There are many surveys of populations that contain a number of extreme values. This is particularly true in surveys of business enterprises. Outliers observations can have an important effect on work with estimation especially on low level of the aggregation. Although the values are extreme, they need not necessarily be false; extremely large observations are a natural component in survey populations. So we shall explore some alternative technique estimation less sensitive to outliers. In this paper we examine local regression which has ability to accommodate local departures from the underlying linear model in business statistics.