Because the field is still developing, there is no single source where one can find everything one needs to know. In fact, there aren’t even completely adequate monographs on any of the most important topics needed for risk assessment. For self study, we recommend the three texts cited in bold below.
Although a bit out of date, Morgan and Henrion (1990) was the first good (and probably still the best) treatment of the elements of uncertainty analysis including sensitivity analysis. Saltelli et al. (2000) is an edited collection on sensitivity analysis. Although quite current, it has a rather limited perspective and does not address the complementary issues of forward analysis. Warren-Hicks and Moore (1998) addresses a surprising number of the important issues in an excellent way, although, because it is an edited collection, it lacks a coherent voice and structure. Vose (1996) and Cullen and Frey (1999) are also important texts, but they have far too many serious mistakes and structural flaws to be useful for self study.
Meyer and Booker (1991) comprehensively reviews the elicitation of expert opinion and its use in formal analyses. Cooke (1991) also surveys the use of expert opinion in risk assessments and contains several important ideas, but its brash approach neglects many of the most basic considerations.
Finkel (1994) offers a short but comprehensive review of the pitfalls of decision analysis, and its essentiality in risk analysis. He shows how the best answer can sometimes depend on the context in which a question is asked or even how it is posed. Kmietowicz and Pearman (1981) provide a short and excellent review of classical decision theory and several important extensions to the classical theory that make it useful with interval information. There are, in fact, several reasonably good books on decision analysis. Kammen and Hassenzahl (1999) is a more popular treatment that is organized for self study and has much to recommend it.
Gigerenzer (2002) is an eye-opening account of the facilities humans use in perceiving numerical and especially probabilistic information. The first chapter is an overview of the entire book. It is extremely easy to confuse and fool people with numbers, including most scientists and virtually every medical doctor lawyer. Yet Gigerenzer’s argument goes far beyond a charge of innumeracy (Paulos 1988); he offers guidance on how to reword probabilistic statements so that they are intelligible to humans. The impact of his simple suggestions is remarkable. This book is based on research that is related to the psychometric literature as excellently summarized in Kahneman et al. (1982).