A Bayesian normal model combines a normal distribution for observations with a prior distribution for its unknown parameters. For independent with known and prior , normal-normal conjugacy gives posterior variance . The posterior mean is the corresponding precision-weighted average of and the sample mean. Unknown observation variance requires an additional prior; treating it as known yields different uncertainty. The model makes prior information and measurement variation explicit.
For iid observations, the prior density relative to gives posterior kernel . It is proper exactly when and . The conditionals are and . Integrating out changes the inverse-gamma shape to .
For independent with known and prior with fixed , the posterior has variance and meanAs , and , so fixed-level posterior credible intervals agree asymptotically with the corresponding known-variance confidence intervals.
For repeated independent measurements with prior and known scales, normal-normal conjugacy gives and . The next measurement has posterior predictive distribution . Its one-sided conditional posterior-predictive false-alarm level yields the displayed limit; this is a per-measurement probability, not a simultaneous false-alarm guarantee across repeated monitoring. Predictive width tends to the irreducible measurement width, while the center adapts to the individual's data. To assess a held-out measurement, fit with earlier measurements only.
For and independent normal observations conditional on , with mean and variances , and normal prior mean and variance , write . Multiplying likelihood and prior adds precisions . completing the square gives the displayed posterior distribution. Its posterior mean is , where and . Thus the Bayes estimator under squared error loss is exactly the Bühlmann–Straub credibility estimate, not merely an approximation to it. Its posterior variance is .
Articles by others on the same topic
There are currently no matching articles.