Locally weighted regression is the method of regression analysis in which the polynomials of degree one (linear) or two (quadratic) are used to approximate regression function in particular 'neighbourhoods' of the space of explanatory variables. It is many times useful for smoothing scatter diagrams to allow any structure to be seen more clearly and for identifying the possible non-linear relationships between the response and the explanatory variables. A robust estimation procedure (which is usually known as loess) is taken in use to guard against deviant points distorting the smoothed points. Essentially the procedure involves an adaptation of the iteratively reweighted least squares. The example shown in the figure illustrates the situation in which the locally weighted regression differs considerably from the linear failure of y on x as fitted by least squares estimation.