Bayesian Inference for Means

Alan Agresti, Maria Kateri, Ranjini Grove et al. · Chapman & Hall/CRC eBooks · 2026

This chapter presents Bayesian methods for estimating means. Standard models use for their sampling model the normal (Gaussian) distribution, which is the conjugate prior distribution for estimating the mean. The conjugate prior distribution for estimating the variance is the inverse gamma distribution, which corresponds to a gamma prior distribution for the precision, its reciprocal. Model-fitting is also possible with other prior distributions, such as an exponential distribution for the variance or improper prior distributions, such as a constant over the real line. We can simulate the posterior distribution by simulating the variance from its posterior distribution, and then given the variance, simulating the mean from the posterior conditional distribution. With the improper prior distributions, posterior intervals for the mean are identical to frequentist confidence intervals using the t distribution. The same prior distributions can be used to compare two means or to make inferences about several means. For large samples, Bayesian estimators share the same good large-sample performance as maximum likelihood estimators, such as being relatively efficient and having approximately normal distributions.

Cited by 0 works. doi:10.1201/9781003715924-3

Authors