Analyzing Dependent Variables with Multiple Surrogates in Accounting
Research
Abstract
The paper contains details a research carried out to show that the use
of geometric mean to unify multivariate dependent variables in financial
performance studies gives better and more practical results than the
multiple abstraction analysis provided using advanced econometric tools
such as TLS, PLS, MCA, Canonical correlations etc.
The study used the logistic regression analysis to compare the a priori
expectations of 30 Ph.D research theses with their actual outcomes using
econometric tools and the actual outcome using geometric means. The
study used Ph.D theses in accounting and finance sourced from the
libraries of four universities in Nigeria.
The study was a desktop research using publicly available literary
resources and as such requires no ethical clearance.