A kind of Markov chain Monte Carlo, where we sample from the conditional posterior distribution with for a parameter, with all other parameters fixed. This is repeated for all parameters. Eventually the MCMC converges to the real posterior.