Statistics for small data
Abstract
The fact that risk assessors often have to rely on “small” data cause tension in our perception of statistical inference in the context of risk analysis. The epistemic situation for risk analysis is seldom ideal for... [ view full abstract ]
The fact that risk assessors often have to rely on “small” data cause tension in our perception of statistical inference in the context of risk analysis. The epistemic situation for risk analysis is seldom ideal for statistical inference in a traditional sense. On the contrary, the aim of risk analysis is to evaluate what we know - and do not know - about a future event, for which we have more or less experience of similar kinds of events. I propose a conceptual framework for predictive inference that aim to understand the meaning of “small” and the consequences thereof for the treatment of uncertainty in predictive inference. The framework distinguish different processes involved in predictive inference i.e. system process, observation process, expert elicitation process, prediction process and conditioning process. I demonstrate the framework on chemical hazard assessment using Species Sensitivity distributions informed by a mixture of testing and non-testing information.
Authors
-
Ullrika Sahlin
(Lund University)
Topic Areas
Evidence to inform risk relevant policy , Decision-making and uncertainty
Session
T5_B » Advances in theory & practice 2 (13:30 - Monday, 20th June, CB3.15)
Presentation Files
The presenter has not uploaded any presentation files.