Journal article
Evaluating Effects of Independent Variables
Political methodology, Vol.3(1), pp.27-47
01/01/1976
Abstract
This paper develops a coefficient, referred to as an "effects coefficient," useful for evaluating the effects of independent variables in any linear causal system. The effects coefficient differs from earlier efforts to derive "total effects" in a number of critical ways. First, the definition of the effect of variable i on variable k as the sum of direct and indirect effects, Eki = DEki + IEki, is formalized and free of dependence on the correlation coefficient. Further, Eki may be calculated by a quite simple algorithm, rather than by the burdensome "basic theorem of path analysis," or by an error-prone, arrow-tracing procedure. Also, the effects coefficient is appropriate for just-identified and for overidentified causal systems. In addition, and perhaps most importantly, the concept of an effects coefficient is useful for nonrecursive, as well as recursive, causal systems. The effects coefficient, Eki, thus emerges as a general summary measure of a variable's effect in a causal system. As such, it is preferable to ordinary correlation or regression coefficients for determining the relative importance of several independent variables.
Details
- Title: Subtitle
- Evaluating Effects of Independent Variables
- Creators
- Michael S. Lewis-BeckLawrence B. Mohr
- Resource Type
- Journal article
- Publication Details
- Political methodology, Vol.3(1), pp.27-47
- Publisher
- Geron-X, Inc
- ISSN
- 0162-2021
- Language
- English
- Date published
- 01/01/1976
- Academic Unit
- Political Science
- Record Identifier
- 9984025520102771
Metrics
8 Record Views