A Comparison: Traditional Canonical Correlation Analysis versus the Mimic/sem Approach Using a Multi- Group Study of Corporate Size versus Profitability

This study employs the use of two methods of canonical correlation analysis to investigate the relationship between two latent variables: corporation “size” and “profitability”. The variables examined are divided into an independent set of variables representing the latent variable “size”, and a dependent set of variables representing the latent variable called “profitability”. The first method employed is the traditional canonical correlation analysis (CCA) approach. The second method used is the structural equation modeling (SEM) approach. The SEM approach uses the Multiple Indicators / Multiple Causes (MIMIC) model. Three variables are used to measure company size; number of employees, total assets, and total income. Three variables are also used to indicate company profitability; return on equity, return on income, and earnings per share. This study demonstrates that the MIMIC model can be quite useful and in some ways is superior to the traditional canonical approach. The traditional approach finds that the first canonical function is significant and the same is found with the MIMIC model. Furthermore, it was shown that multiple groups which are moderated by categorical variables, in this case industry sector, can be evaluated for differences using the MIMIC approach. The traditional model does not allow this. In the literature review no other study was found which demonstrates the use of the SEM/MIMIC model in the examination of the affects of categorical moderators.