TABLE OF CONTENTS Ten requirements that an empirical modeling methodology should satisfy have been expounded, explained and demonstrated in Chapter 6. Current major methodologies for empirical modeling were then assessed for compliance with these requirements. In particular, existing major methodologies for relational modeling, namely, linear regression, data normalization and generalized linear models (GLM) were addressed. We have realized that these methodologies fall short of satisfying the desirable requirements.
In previous chapters of this part of the book (Chapters 7 12), the RMM approach has been developed, the model, its variations and fitting and estimation procedures. We have also demonstrated that numerous current models of both systematic variation and random variation are indeed special cases of the RMM model.
It is now appropriate to subject the RMM approach to the same scrutiny that current methodologies have undergone in Chapter 6.
Evaluating RMM for compliance to requirements is partitioned into two, similarly to earlier evaluations: Compliance to requirements relevant to modeling systematic variation (Section 13.2) and compliance to those relevant to modeling random variation (Section 13.3).
We compare in this section the RMM properties relative to the two major alternatives for modeling systematic variation, namely, the BC normalizing transformation and GLM. Linear regression is a special case of both and therefore will not be addressed separately.
Evaluation of RMM relative to the requirements, as expounded in Chapter 6, is displayed in a tabular form in Table 13.1. This is an expansion of Table 6.1, which now includes...
