Update section3_mcmc.jl

section3_mcmc.jl

@@ -287,7 +287,7 @@ A class of methods called **Markov Chain Monte Carlo** (MCMC) is a popular and s
   * **Monte Carlo**: estimation by the Markov chain samples as illustrated in the previous section
 
 
-The idea is to produce posterior samples ``\{\theta^{(r)}\}_{r=1}^R`` **in sequence**, each one depending only on ``\theta^{(r-1)}`` and not on its more distant history of predecessors, i.e. a **Markov Chain** (which accounts for the first **MC** of the acronym **MCMC**). When the transition probability of the chain satisfies certain conditions, Markov Chain theory then says that, under quite general conditions, the empirical distribution of the simulated samples will approach the desired target distribution as we simulate the chain long enough, i.e. when ``R`` is large.
+The idea is to produce posterior samples ``\{\theta^{(r)}\}_{r=1}^R`` **in sequence**, each one depending only on ``\theta^{(r-1)}`` and not on its more distant history of predecessors, *i.e.* a **Markov Chain** (which accounts for the first **MC** of the acronym **MCMC**). When the transition probability of the chain satisfies certain conditions, Markov Chain theory then says that, under quite general conditions, the empirical distribution of the simulated samples will approach the desired target distribution as we simulate the chain long enough, i.e. when ``R`` is large.
 
 Since the sequence converges to the target distribution when ``R`` is large, we usually only retain the last chunk of the sequence as "good samples" or equivalently **discard** the initial samples as **burn-in** since they are usually not representative of the distribution to be approximated. For example, if the Markov Chain has been simulated by 4,000 steps, we only keep the last 2000 to form an empirical distribution of the target.
 """