My model involves solving an equation of the F(d, b e^{k}) = \frac{q}{c - 1} for b, where F(.) is the gamma CDF. Obviously, b = \frac{F^{-1}(\frac{q}{c - 1}, 1, d)}{e^{k}} , where F^{-1} is the inverse of the incomplete gamma quantile function. Is there a way to implement this in stan? The func…

Alright, so for the function: F^{-1}\left(x; \alpha, \beta \right) = \frac{Q^{-1}(\alpha,1-x)}{\beta} Which is given in Wolfram Alpha by InverseGammaRegularized[a,1-x]/b, and can be verified by comparing against qgamma(x,shape=a,rate=b). The partial derivatives for each input are given by: x:…

[image] andrjohns: #include <boost/math/special_functions/hypergeometric_pFq.hpp> #include <stan/math/prim/scal.hpp> #include <stan/math/rev/core.hpp> // Function to handle Stan parameters inline stan::math::var gamma_icdf(const stan::math::var& x, const stan::…

Can you post the stan_model call that you’re using and the result of calling cat on the Stan and C++ files?

So after some more testing, it looks like including the Boost header (hypergeometric_pFq.hpp) is causing issues with rstan’s usage of other boost functions. The best way forward here is to use cmdstanr: R Interface to CmdStan • cmdstanr , which includes external c++ in a different way. Once you hav…

After several attempts I finally have cmdstan and cmdstanr installed. I tried running the codes below but ended up with Error 127 mingw32-make.exe: *** [make/program:47: C:\Users\lucp9544\AppData\Local\Temp\Rtmpa05r64\model-4714e0f2622.hpp] Error 127. I tried the bernoulli example which works fine. …

What happens if you run the script I posted above with no changes?

[image] andrjohns: library(cmdstanr) hpp = "#include <stan/math/prim.hpp> #include <stan/math/rev/core.hpp> #include <boost/math/special_functions/hypergeometric_pFq.hpp> // Function to handle Stan parameters inline stan::math::var gamma_icdf(const stan::math::var& x, …

Hmm, @spinkney would you mind running the above script to see if it works for you? Just to see whether it’s only my installation that it works for

Hello @andrjohns , Any update? I still tried again with no success. If I run the model without the C++ file, it works just fine. Do you also use a windows system?

Good timing! I was just looking into this. It looks like the error has something to do with how cmdstanr handles Windows paths. I can replicate your error, and then fix it if I use the Windows short path name. Can you try the cmdstan_model call again, but this time wrap the getwd() calls in a call …

Gamma distribution quantile and inverse quantile function

General

andrjohns June 14, 2021, 1:23pm 40

Here’s a gamma_icdf that has overloads for all combinations of data and parameter:
gamma_icdf.hpp (9.3 KB)

3 Likes

Converting Stan data types to C/C++ primative types

Topic		Replies	Views
Gamma quantile function Developers features	29	5123	July 1, 2020
Boost function in rstan RStan	9	1663	July 1, 2020
Possible Issue with gamma_lccdf Developers	22	643	February 6, 2026
Quantile function of gamma distribution Modeling rstan , techniques , fitting-issues , specification , math	1	527	September 15, 2021
Using integration to fit the difference of gamma distributed random variable Modeling fitting-issues , integrator , user-defined-functions	15	1588	July 4, 2020

Gamma distribution quantile and inverse quantile function

Related topics