Multiple Regression in Practice

Abstract

Multiple regression analysis is one of the social sciences most popular procedures. This monograph provides a systematic treatment of many of the major problems encountered in using regression analysis. Because it is likely that 1 or more of the regression models assumptions will be violated in a specific empirical analysis the ability to recognize problems and take appropriate action helps to ensure the proper use of the technic. The author clearly and concisely discusses the consequences of violating the assumptions of the regression model procedures for figuring out when violations exist and strategies for dealing with problems when they occur. Problems of specification error measurement error multicollinearity nonlinearity and nonadditivity and heteroscedasticity and autocorrelation are discussed with many examples from political science sociology and economics. Because many applications of regression in the social sciences involve analysis of causes randomly drawn from a larger population some of the major examples are constructed to show more clearly the properties of regression estimates derived from samples. The concepts of bias and efficiency in statistical estimation receive particular attention.