Extrapolation and Effects of Outliers on Regression Line

Extrapolation(외삽법)

Making predictions outside the range of data provided by using statistical calculations(usually regression line).

Extrapolation should be avoided if possible.

 

Effect of outliers in regression

Different types of outliers have different effect on regression line.

-       If the outlier’s x value is outside the cluster’s x value range, it is called outlier in the x direction.

-       If the outlier’s y value is outside the cluster’s y value range, it is called outlier in the y direction.

-       If the outlier’s x and y value is within the cluster’s range but falls outside the pattern of data, it is called ‘bivariate outlier’.

Outliers in the x direction greatly influence the position of the regression line.

Outliers in the y direction barely affect the position of the regression line.

If the outlier is on the regression line, it doesn’t affect the regression line.

Bivariate outlier doesn’t affect the regression line because it is not a outlier in the x or y direction.

Outlier that affects the regression line greatly is called ‘Influential outlier’. Outlier in the x direction has a higher chance to be an influential outlier.