Examples of these algorithms include Gibbs sampling and the Metropolis–Hastings algorithm. For 10 points each:
[10m] Name this class of sampling methods typically referred to by a four-letter acronym. These methods aim to sample from the stationary distribution of a memoryless stochastic process.
ANSWER: Markov chain Monte Carlo [or MCMC; prompt on Markov chain or Markov processes; prompt on Monte Carlo methods]
[10e] This statement for MCMC models guarantees convergence to the target distribution. More generically, this statement states that the rescaled distribution of sample means converges to a normal distribution.
ANSWER: central limit theorem [or CLT]
[10h] Charlie Geyer critiqued this procedure as unnecessary since it appeals to a “central limit almost-but-not-quite theorem for almost-but-not-quite stationary processes.” This procedure involves initially running the Markov chain for n steps, and then throwing away the results.
ANSWER: burn-in
<GC, Other Science>