The subject of this book is the reasoning under uncertainty based on statistical evidence. The concepts are developed, explained and illustrated in the context of the mathematical theory of hints, which is a variant of the Dempster-Shafer theory of evidence. In the first two chapters, the theory of generalized functional models for a discrete parameter is developed, which leads to a general notion of weight of evidence. The second part of the book is dedicated to the study of special linear functional models called Gaussian linear systems. Finally, it is shown that the celebrated Kalman filter can easily be derived by local propagation of Gaussian hints in a Markov tree. The subject of this book is the reasoning under uncertainty based on sta tistical evidence, where the word reasoning is taken to mean searching for arguments in favor or against particular hypotheses of interest. The kind of reasoning we are using is composed of two aspects. The first one is inspired from classical reasoning in formal logic, where deductions are made from a knowledge base of observed facts and formulas representing the domain spe cific knowledge. In this book, the facts are the statistical observations and the general knowledge is represented by an instance of a special kind of sta tistical models called functional models. The second aspect deals with the uncertainty under which the formal reasoning takes place. For this aspect, the theory of hints [27] is the appropriate tool. Basically, we assume that some uncertain perturbation takes a specific value and then logically eval uate the consequences of this assumption. The original uncertainty about the perturbation is then transferred to the consequences of the assumption. This kind of reasoning is called assumption-based reasoning. Before going into more details about the content of this book, it might be interesting to look briefly at the roots and origins of assumption-based reasoning in the statistical context. In 1930, R. A. Fisher [17] defined the notion of fiducial distribution as the result of a new form of argument, as opposed to the result of the older Bayesian argument. 1. The Theory of Generalized Functional Models.- 2. The Plausibility and Likelihood Functions.- 3. Hints on Continuous Frames and Gaussian Linear Systems.- 4. Assumption-Based Reasoning with Classical Regression Models.- 5. Assumption-Based Reasoning with General Gaussian Linear Systems.- 6. Gaussian Hints as a Valuation System.- 7. Local Propagation of Gaussian Hints.- 8. Application to the Kaiman Filter.- References.
From the reviews:
"The book, clearly written, is structured in two parts comprising two and six chapters, respectively. ... The overall quality of the book is very good, the material is well organized and notations and terminology are unified. It is a very good presentation of the state of research in the area of modeling reasoning and an insightful reference for the statistician." (Evdokia Xekalaki, Zentralblatt MATH, Vol. 1100 (2), 2007)
