Bootstrapping

NinaJk

Hi everyone,

I was wondering if a bayesian variant of bootstrapping (for a moderation) is possible in JASP? 

Thank you in advance

EJ

Not yet, but you could issue a feature request on out GitHub page. You might want to explain why you want the Bayesian bootstrap though (in this case you can have MCMC samples from the posterior distribution).

Cheers,

E.J.

NinaJk

Thanks for your response!

The reason I wanted to use bootstrapping was because of the assumptions that were violated and the small sample size. But if I understand correctly, I can better use the MCMC samples from the posterior distribution (and I think therefore I have to look at the credible interval the same way you look at the confidence interval when you use bootstrapping?)

EJ

Hi NinaJk,

I'm not sure bootstrapping will help with model misspecification or small sample sizes -- the bootstrap was meant to obtain an estimate of sampling variability in cases where no analytic solution is available (which includes cases where there is a solution, but one that is asymptotic only). Credible intervals are often numerically similar to confidence intervals, but their interpretation is subtly different. Basically, credible intervals are what most people believe are confidence intervals :-) See for instance: https://link.springer.com/article/10.3758/s13423-015-0947-8

Cheers,

E.J.