Timeline for Choosing between mean, median, and mode as estimator
Current License: CC BY-SA 3.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 3, 2017 at 14:43 | comment | added | Dave Harris | I would run a kernel density estimate on your parameter estimates and see what they really look like. My guess is that you are going to find a surprise in there somewhere. | |
| May 3, 2017 at 14:25 | comment | added | hipHopMetropolisHastings | In response to your edit, that parameters I am constraining are the covariance matrix, and the Markov chain transition probabilities. Only conditional on the transition probabilities can I assume the likelihood is normal. Optimizing across co-variance matrix I believe allows to get the correct constant of integration. | |
| May 3, 2017 at 1:42 | history | edited | Dave Harris | CC BY-SA 3.0 |
Respond to edit
|
| May 2, 2017 at 20:10 | comment | added | hipHopMetropolisHastings | Thanks for the answer. I don't have the reputation to upvote but it was very helpful in my interpretation. Obviously I am seeing asymmetric confidence intervals. My guess is this is because I am constraining some variables. For example, if I constrain the estimation of $\beta \in (0,1)$, and the true $\beta = 0.99$, then obviously there is a lot more room to move around for values less than 0.99, resulting in bias of the mean. Does this intuition have theoretical underpinnings? | |
| May 2, 2017 at 19:52 | history | answered | Dave Harris | CC BY-SA 3.0 |