Errors on MAP

[benrouse]

Hi,

I was wondering if there was anything built into prospector to give the errors on the MAP?

So far, I have been plotting the SED with the percentiles and MAP, but sometimes I get plots like the one below, where the MAP fits, but the percentiles and median are completely off.

Thanks in advance.

[bd-j]

Are you taking percentiles of the individual parameters and then constructing SEDs from those? That will not work, since in the presence of parameter correlations the model described by the (say 50th) percentile of each parameter is not at all guaranteed to look anything like the data (think of a posterior that looks like a banana in two parameters; using the posterior median of each parameter would put you in a part of parameter space where there is no posterior probability!)

To get SED percentiles you have to draw some number of posterior parameter positions fairly from the chain, construct the predicted spectrum for each one, and then take the percentiles of the fluxes at each wavelength.