The Well-Calibrated Bayesian

By A. P. Dawid. 1982

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

On Fairness and Calibration (2017, methodology): “When risk tools are used in practice, a key goal is that they are calibrated: if we look at the set of people who receive a predicted probability of p, we would like a p fraction of the members of this set to be positive instances of the classification problem [11].”
BatchEnsemble: An Alternative Approach to Efficient Ensemble and Lifelong Learning (2020, background): “Several measures have been proposed to assess the quality of uncertainty estimates, such as calibration (Dawid, 1982; Degroot & Fienberg, 1983).”
Measuring Calibration in Deep Learning (2019, background): “One mathematical formulation of the reliability of confidence is calibration (Murphy and Epstein, 1967; Dawid, 1982).”
Bayesian Inference of Sampled Ancestor Trees for Epidemiology and Fossil Calibration (2014, background): “e, the median of the estimated medians, the median of the relative errors and so forth) and count the number of times when the true value belongs to the 95% HPD interval [43].”
Beyond subjective and objective in statistics (2015, methodology): “There is also some work on (partial) intersubjective agreement on prior specifications, e.g. Dawid (1982a), providing a rationale for consensus (virtue V2(c)).”
Any Inspection is Manipulable (2001, methodology): “The forecaster wins the game if µ passes the calibration test induced by D at ω∞, and looses otherwise.”
Diagnostics of prior-data agreement in applied Bayesian analysis (2008, background): “Main references are De Finetti [26], Dawid [25], Hill [39], Lucas [42] and especially O’Hagan [43–45,48], Andrade and O’Hagan [3] and Angers [5].”
Calibrated Bayes (2006, background): “Many statisticians have advanced the calibrated Bayesian idea; some examples are Peers (1965), Welch (1965), and Dawid (1982). I myself have assessed the frequentist properties of Bayesian procedures in methodological work (e.g., Little 1988). But two seminal papers by leading proponents of this sch”
Heuristics and Biases: The Weighing of Evidence and the Determinants of Confidence (2002, unknown): “This objection, however, overlooks the fact that a Bayesian expects to be calibrated (Dawid, 1982), hence the theory of subjective probability permits the comparison of confidence and accuracy.”
The Implicit Fairness Criterion of Unconstrained Learning (2018, background): “The concept of calibration has a venerable history in statistics and machine learning (Cox, 1958; Murphy and Winkler, 1977; Dawid, 1982; DeGroot and Fienberg, 1983; Platt, 1999; Zadrozny and Elkan, 2001; Niculescu-Mizil and Caruana, 2005).”
Prediction, learning, and games (2006, background): “It was Dawid [80] who introduced the problem to the literature of mathematical statistics – see also Dawid’s related work on calibration and “prequential” statistics [81, 82, 84].”
Testing Multiple Forecasters (2008, unknown): “Thus we modify the inequality in (3) to a strict inequality (on one side of the interval) when needed, to reflect that the empirical frequency must converge to the appropriate interval.”
bModelTest: Bayesian phylogenetic site model averaging and model comparison (2015, background): “On average, one would expect the coverage to be 95% if simulations are drawn from the prior [25], so each entry in Table 1 has an expected value of 95, but can deviate due to small sample size.”
Non-Bayesian Testing of a Stochastic Prediction (2006, methodology): “The existing literature focused on calibration tests (Dawid (1982, 1985)).”
The Power of Ensembles for Active Learning in Image Classification (2018, background): “Uncertainty calibration To assess calibration ([7]) quality we determine whether the expected fraction of correct classifications (as predicted by the model confidence, i.”
Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles (2016, methodology): “Firstly, we shall examine calibration [12, 13], a frequentist notion of uncertainty which measures the discrepancy between subjective forecasts and (empirical) long-run frequencies.”
Fairness in Machine Learning: A Survey (2020, background): “Synopsis: Calibration is the process of ensuring that the proportion of positive predictions is equal to the proportion of positive examples [83].”