Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Monte Carlo standard error for posterior predictive p values #5

Open
crsh opened this issue Jun 29, 2018 · 2 comments
Open

Monte Carlo standard error for posterior predictive p values #5

crsh opened this issue Jun 29, 2018 · 2 comments
Assignees

Comments

@crsh
Copy link
Contributor

crsh commented Jun 29, 2018

Hi. I was wondering whether we want to plot the Monte Carlo standard error or confidence interval for the estimated posterior predictive p values (PPP) in the summary plot. This would give us a sense of how reliable that estimate is in terms of a dichotomous decision about model fit (as this is often how this is used in practice). I'm happy to add this. All I would need is the effective sample size for the sampling statistic used to calculate the PPP. Is this stored somewhere in the MPTmultiverse objects?

@danheck
Copy link
Collaborator

danheck commented Jun 30, 2018

This information is currently not available in the fitted object. The MCMC standard errors are currently only available for the parameters.

This is also due to the fact that TreeBUGS does not provide SEs for the p-values as output. However, it might be worth adding that (however, I have not yet seen this in any paper). How exactly would you compute the SE? Simply by: SE(ppp) = sqrt(ppp*(1-ppp) / n_eff) ?

@crsh
Copy link
Contributor Author

crsh commented Jul 2, 2018

Yes, as per p. 267 in Gelman, et al. (2015). The confidence interval could then be calculated either asymptotically by relying on CLT or as upper bound by multiplying the asymptotic interval by 2.3 (Rosenthal, J. S. (2017). Simple confidence intervals for MCMC without CLTs. Electronic Journal of Statistics, 11(1), 211–214. https://doi.org/10.1214/17-EJS1224). Do you happen to know whether it's sensible to rely on CLT in this case?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants