[R-sig-ME] Interpreting mcmcsamp results

[Ben Bolker]

hits=-2.6 tests=BAYES_00,SPF_HELO_PASS
X-USF-Spam-Flag: NO


Kevin E. Thorpe wrote:
> When I run summary (from the coda package) on the result of mcmcsamp I
> get, among other things, some quantiles for the model parameters fit in
> my lmer() call.
> 
> Is it correct to interpret say the 2.5% to 97.5% as a 95% credible interval?
> 
> Thanks,
> 
> Kevin
> 


    Since I haven't seen any other answers I will take a crack at
this.

    coda:::summary.mcmc (which is the function that gets called when
you ask for a summary of an mcmc) gives quantiles, not a credible
interval.  If you want credible intervals, try HPDinterval from
the coda package (which will give 95% credible intervals by default).
For symmetric distributions the credible interval and the quantile
should be the same.

    Ben Bolker