Multiple Regression Models II: Crucial Extensions
Paul M. Kellstedt, Guy D. Whitten · Cambridge University Press eBooks · 2008
OVERVIEW In this chapter we provide introductory discussions of and advice for commonly encountered research scenarios involving multiple regression models. Issues covered include dummy independent variables, interactive specifications, dummy dependent variables, influential cases, multicollinearity, and models of time-series data. EXTENSIONS OF OLS In the previous two chapters we discussed in detail various aspects of the estimation and interpretation of OLS regression models. In this chapter we go through a series of research scenarios commonly encountered by political science researchers as they attempt to test their hypotheses within the OLS framework. The purpose of this chapter is twofold – first, to help you to identify when you have hit these issues and, second, to help you to figure out what to do to continue on your way. We begin with a discussion of “dummy” independent variables and how to properly use them to make inferences. We then discuss how to test interactive hypotheses with dummy variables. Our third topic with dummy variables involves the interpretation of models in which our dependent variable is a dummy variable. We next turn our attention to two frequently encountered problems in OLS – outliers and multicollinearity. With both of these topics, at least half of the battle is identifying that you have the problem. Finally, we conclude with a discussion of a series of problems specific to the analysis of time-series data. BEING SMART WITH DUMMY INDEPENDENT VARIABLES IN OLS In Chapter 6 we discussed how an important part of knowing your data involves knowing the metric in which each of your variables is measured.