Bayesian inference


Bayesian inferencesearch for term

Likelihood is the probability of the data given the model; Bayesian inference instead deals with the probability of the model given the data, also named ‘posterior probability’. This posterior probability of a model is proportional to the product of the likelihood and of a ‘prior probability’. Such a prior probability permits incorporation of exterior knowledge into an analysis; for instance, one could assume that the prior probability over the transition transversion ratio in a particular dataset follows a uniform distribution on. Contrary to Maximum Likelihood inference, the common practice in Bayesian inference is not to return parameter values of the highest posterior probability; instead, whole distributions of parameter values are returned. To obtain these distributions, MCMC techniques are often used. (Bousseau 2009)