ABSTRACT
In this chapter we consider a wide range of aspects of model-fitting which have not yet been discussed. In most cases, we develop general principles, such as the δ-method, a test for parameter-redundancy, the EM (expectationmaximisation) algorithm, etc. We also discuss the breakdown of the customary regularity conditions, and present several methods of parameter-estimation which are alternatives to the method of maximum likelihood, such as the method of minimum chi-square, and the method of empirical transforms. The Bayesian approach to model-fitting is so widely applicable, that we defer discussion of this important approach until Chapter 7, which is devoted entirely to Bayesian statistics and modern computational methods for conducting Bayesian inference.
