Steven Murray

Hmm, well that would make sense! But as far as I can tell, my token points to a secret that definitely exists. Unless it's case-sensitive.

Thanks @Ipgera that worked for me too. Probably this should be added to the docs...

The generalised gamma distribution is defined [here](https://en.wikipedia.org/wiki/Generalized_gamma_distribution). As you can see, all three of its parameters are restricted to be positive. In particular, if $d$ here is negative, the distribution...

Sorry about the math... apparently there is no support for latex in GFM

@bob-carpenter The trick will be if the included incomplete gamma gradient has support for negative d.

There is a workaround to this in some cases -- for small negative d, a recurrence relation can be used. See my own answer to the question at http://stats.stackexchange.com/questions/161134/weighted-log-probabilities-in-generalised-gamma-distribution/162390#162390 Might...

I've done a bit of work using my "fast" function, which only calculates the gamma function, gammainc(a,x), for a>-2. For reference, the function block is: ``` functions { /** *...

Okay, so I've re-implemented in a more general manner, so that it works for any negative a. I also took your suggestions into account, so here is the code: ```...

Just had a quick read-up of binary searches on Wikipedia and I think we'd have to do a one-sided binary search starting at 0, which has performance of 2*log2(n). So...

Cool, I'll keep it as-is for now. By the way, is there any way for the user to specify derivatives/jacobians? I would have thought that if there was a way,...