What diagnostic tools do you, AG, recommend for assessing the information contained about a particular parameter, or set of parameters, in the data?
What tools do you, AG, recommend for designing efficient data collection procedures based on a desire to learn about specific sets of parameters (e.g., under what conditions will the posterior mean/mode be a biased estimator of the "true" parameter?" - might not just be a sampling issue, in that the same bias is easily repeatable in estimation that uses 3000 sample time points for each of 500 replicates that resolve the state space pretty well... but maybe there is still an issue in sample design)
o how can we determine whether it's better to make observations of trajectories for longer periods of time vs. at higher resolution time scales if we want to learn about a given parameter)?
Reminder to ask Andrew Gelman:
o how can we determine whether it's better to make observations of trajectories for longer periods of time vs. at higher resolution time scales if we want to learn about a given parameter)?