Question about plotting grid scans using SB

[Tom Varley]

So I used SuperBayes to do an ordinary, standard grid scan for
comparative purposes and was wondering what would be the best way in
general to plot it?

The standard output from SB will be a posterior plot and a likelihood
plot, both of which assign weights/probabilities which in the case of
a grid scan I'm guessing are not correct/relevant. My solution was to
take the likelihood plot and assign all points with non-zero
probability with an equal importance (i.e all the same color), is this
an acceptable approach?

I much appreciate any help given!

[Roberto Trotta]

> The standard output from SB will be a posterior plot and a likelihood
> plot, both of which assign weights/probabilities which in the case of
> a grid scan I'm guessing are not correct/relevant. My solution was to
> take the likelihood plot and assign all points with non-zero
> probability with an equal importance (i.e all the same color), is this
> an acceptable approach?

The chain will contains only the acceptable (ie, physical) points, all
with weight 1 (as in this case there is no probabilistic significance
to the points). The only relevant quantitity is the log like, as you
say.

I'm not sure what you mean precisely by "standard output" - is this
after you processed the chains using getplots? For chains produced
with grid scan I would advice not to use getplots at all, as the
posterior plot would be meaningless. The average like might me more of
a useful quantity to look at, but again would have no statistical
significance. Better to either plot the points in relevant 2D planes
by hand from the chains directly, or simply select the highest
likelihood along the hidden dimensions - this is somewhat analogous to
profile likelihoods. I'm not sure your procedue can be easily
interpreted statistically. It all depends what are you trying to
deduce from the scan...

Hope this helps, let me know if you have further questions.

Roberto