Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Supported by

Impossible values included in credible interval

Hey,


I am preforming a mixed BANOVA with RMSE as the dependent variable. I noticed that in the descriptives the credible intervals include negative values (which are of course impossible). It seems that the credible interval is centered around the mean, even when all deviations from the mean are positive. Is it possible to modify the analysis so that credible intervals do not extend below 0? What setting is in that case needed? How come the credible interval centers on the mean even though deviations from the mean are skewed?


Best,

August

Comments

  • Hi August

    Because it's descriptives, and not inference, we use the credible intervals with uniform priors on the entire real line (so basically a confidence interval). I've always been uneasy with this; maybe it would be best to leave the user the option on having the intervals from the model, or from the uniform. You could indicate something like this on our GitHub page.

    Cheers,

    E.J.

  • Hi E.J,


    Ok, then my intuition was correct, that it was basically a confidence interval. I seems like a bit of waste to not make intervals from the model, since one then misses one of the perks with bayesian stats (having intervals that makes sense). I'll post this as a suggest on the GitHub page.

    //August

Sign In or Register to comment.