Markov chain Monte Carlo with PyMC
Markov Chain Monte Carlo (MCMC) is a technique for generating a sample from a distribution, and it works even if all you have is a non-normalized representation of the distribution. Why does a data scientist care about this? Well, in a Bayesian analysis a non-normalized form of the posterior distribution is super easy to come by, being just the product of likelihood and prior - so MCMC can be used to sample from (essentially simulate) a Bayesian posterior. In python one of the most widely used packages for doing exactly this is called PyMC . What is Markov chain Monte Carlo A Markov Chain is a sequence of RVs {X} each of which will have an observed value from the state space of possible values, {x}. A Markov Process is a sequence of such RVs where the distribution of the initial RV's value is specified (Π0) as well as a Transition Rule (P) which gives the probability to transition from one state to another, P(i,j)=P(Xn+1=xj|Xn=xi), for all pairs of states. Notice P doe