Please use this identifier to cite or link to this item:
http://hdl.handle.net/2122/9261
Authors: | Marzocchi, W.* Jordan, T. H.* |
Title: | Testing for ontological errors in probabilistic forecasting models of natural systems | Journal: | Proceedings of National Academy of Sciences | Series/Report no.: | /111 (2014) | Issue Date: | 2014 | DOI: | 10.1073/pnas.1410183111 | Keywords: | Bayesian statistics testing hazard models |
Subject Classification: | 05. General::05.01. Computational geophysics::05.01.04. Statistical analysis | Abstract: | Probabilistic forecasting models describe the aleatory variability of natural systems as well as our epistemic uncertainty about how the systems work. Testing a model against observations exposes ontological errors in the representation of a system and its uncertainties. We clarify several conceptual issues regarding the testing of probabilistic forecasting models for ontological errors: the ambiguity of the aleatory/epistemic dichotomy, the quantification of uncertainties as degrees of belief, the interplay between Bayesian and frequentist methods, and the scientific pathway for capturing predictability. We show that testability of the ontological null hypothesis derives from an experimental concept, external to the model, that identifies collections of data, observed and not yet observed, that are judged to be exchange- able when conditioned on a set of explanatory variables. These conditional exchangeability judgments specify observations with well-defined frequencies. Any model predicting these behaviors can thus be tested for ontological error by frequentist methods; e.g., using P values. In the forecasting problem, prior predictive model checking, rather than posterior predictive checking, is desir- able because it provides more severe tests. We illustrate experi- mental concepts using examples from probabilistic seismic hazard analysis. Severe testing of a model under an appropriate set of experimental concepts is the key to model validation, in which we seek to know whether a model replicates the data-generating pro- cess well enough to be sufficiently reliable for some useful pur- pose, such as long-term seismic forecasting. Pessimistic views of system predictability fail to recognize the power of this method- ology in separating predictable behaviors from those that are not. |
Appears in Collections: | Article published / in press |
Files in This Item:
File | Description | Size | Format | Existing users please Login |
---|---|---|---|---|
PNAS_Marzocchi_Jordan_2014.pdf | 1.05 MB | Adobe PDF |
WEB OF SCIENCETM
Citations
42
checked on Feb 10, 2021
Page view(s) 50
162
checked on Apr 20, 2024
Download(s)
21
checked on Apr 20, 2024