ABSTRACT
Regression analysis, particularly linear regression, is widely used in engineering, both for statistical prediction purposes and for curve fitting to scattered data. In many cases, the assumptions upon which the analysis method is built are not recognised and misleading or erroneous conclusions are drawn. This paper considers the use of regression in soils engineering, with particular emphasis on problems special to this discipline including selection of independent variables, spurious self-correlation and autocorrelation, potential errors in extrapolation and algebraic manipulations of the regression equations and common mistakes in curve fitting.
