Total factor efficiency/productivity ratio fitting as an alternative to regression and canonical correlation models for performance data

Abstract

This paper discusses a class of modeling alternatives to regression or canonical correlation when dependent variables can be logically considered as outputs to be maximized. Likewise independent variables should be considered as constraints on resources which establish limits to the output levels. A total factor productivity/efficiency ratio of non-negatively weighted outputs divided by similarly weighted inputs is to be fitted to the data by the Maximum Decisional Efficiency Principle. It is assumed that such data, when obtained from experienced managers or viable organizations, should tend to exhibit purposeful rather than random behavior under appropriate parameter value choices and density assumptions. Some model quality improvement issues, analogous to those in regression theory, are also proposed (e.g. criterion choice, residual analysis, and outliers). Potential advantages of the approach are discussed for empirical studies in Information Technology and Production/Operations Management settings.