Functions for simulating a reinforcement learning model in a restless bandit task and estimating its parameters with a variety of methods.
Both R code and Matlab code simulate the same task and estimate the generative parameters with Maximum Likelihood, or Maximum A Posteriori. R code additionally can estimate with Markov Chain Monte Carlo (requires stan). R code also fits lag-1 single and multilevel logistic regression models.
for details:
- R: see Rcode/fitData.R
- Matlab: see matlab/recoverParams.m