Path: Janda's Home Page > Workshop > Canonical Analysis
Canonical Analysis:
"Multiple Regression" between Two Sets of Variables

Kenneth Janda, Northwestern University, Winter Quarter 2003


Virtually all political researchers engaged in multivariate statistical analysis know multiple regression, which computes the best linear combination of a set of independent variables (in a least-squares sense) for predicting to a single dependent variable.

Very few political researchers seem to know about canonical analysis, which extends multiple regression to finding the best linear combination(s) between two sets of variables.

This last session of the workshop will describe canonical analysis and illustrate its application to political research in an article that I wrote with our colleague, Tyler Colman: "Effects of Party Organization on Performance during the 'Golden Age' of Parties," Political Studies, 46 (1998), 611-632.

Outline of Talk

Relevant web sites on the method:
Application and Discussion
Applications in political research:
Rummel's Statistics of Democide
Likelihood of War