– MCMC methods allow us to estimate the shape of a posterior distribution in case we can’t compute it directly.
– Monte Carlo simulations are just a way of estimating a fixed parameter by repeatedly generating random numbers.
– Assuming we have to calculate the area of a circle within the square box. By having points that belong to the circle, we can find out it’s area. Since the circle is inside a square with 10 inch sides, the area can be easily calculated as 78.5 square inches. Instead, however, we can drop 20 points randomly inside the square. Then we count the proportion of points that fell within the circle, and multiply that by the area of the square. That number is a pretty good approximation of the area of the circle.
– Monte Carlo simulations aren’t only used for estimating the area of difficult shapes. By generating a lot of random numbers, they can be used to model very complicated processes. In practice, they’re used to forecast the weather, or estimate the probability of winning an election.
– An important feature of Markov chains is that they are memoryless: everything that you would possibly need to predict the next event is available in the current state, and no new information comes from knowing the history of events.
– Thus, even interdependent events, if they are subject to fixed probabilities, conform to an average.
– More importantly, this prediction isn’t affected at all by which room the person began in! Intuitively, this makes sense: it doesn’t matter where someone is in the house at one point in time in order to simulate and describe where they are likely to be in the long-term, or in general.
– Purpose is to find a fitting posterior distribution. Now in MCMC , we just want to take into consideration a posterior directly that is random has no definite shape or parameters whatsoever. We know that the posterior distribution is somewhere in the range of our prior distribution and our likelihood distribution, but for whatever reason, we can’t compute it directly. Using MCMC methods, we’ll effectively draw samples from the posterior distribution, and then compute statistics like the average on the samples drawn.
– To begin, MCMC methods pick a random parameter value to consider. The simulation will continue to generate random values (this is the Monte Carlo part) , but subject to some rule for determining what makes a good parameter value. The trick is that, for a pair of parameter values, it is possible to compute which is a better parameter value, by computing how likely each value is to explain the data, given our prior beliefs. If a randomly generated parameter value is better than the last one, it is added to the chain of parameter values with a certain probability determined by how much better it is (this is the Markov chain part).
– Since the random samples are subject to fixed probabilities, they tend to converge after a period of time in the region of highest probability for the parameter we’re interested in.
– After convergence has occurred, MCMC sampling yields a set of points which are samples from the posterior distribution.
– Any statistic calculated on the set of samples generated by MCMC simulations is our best guess of that statistic on the true posterior distribution.
– MCMC methods can also be used to estimate the posterior distribution of more than one parameter (human height and weight, say). For n parameters, there exist regions of high probability in n-dimensional space where certain sets of parameter values better explain observed data.