Carlos Parada
Carlos Parada
> > > Thank you so much for the detailed review! I'll pour over these and make the changes as needed thank you so much for the feedback and explanations....
That's extremely weird. Do you know what's causing the performance difference?
Hmm, do you have a [flamegraph](https://timholy.github.io/FlameGraphs.jl/stable/)?
Ahh, I was suggesting you might want to look at the flamegraphs to see which lines specifically are the ones slowing down `logpdf`, sorry for not being clear about that....
> Is it possible that computing the squared Mahalanobis distance for one vector at a time (which is what `logpdf` is doing) is slower than doing it for a matrix...
> Most of the time seems to be taken up by `sqmahal`. Is it possible that computing the squared Mahalanobis distance for one vector at a time (which is what...
> Hmm I see the problem but I still think that a sampler should only return valid samples (or throw an error), but not yield samples that are not in...
> As mentioned e.g. in [#1003 (comment)](https://github.com/JuliaStats/Distributions.jl/issues/1003#issuecomment-555992757), the mass between 0 and `floatmin(Float64)` can be non-negligible. Right, but is there anything we can do about that mass? We can't return...
> I was going to write down the proof because I don't see it written explicitly anywhere. Yeah, I think it's a well-known enough result that we can use it....
> It says, `counts are out of Confidence Interval`. Where are we getting the values of Distribution parameters from? I am not able to reproduce the same for custom values....