I was doing some numerical tests where I ended up comparing rng results from this package by using the characteristic function of the target distribution. At least for alpha in (1,2), the results here do not coincide with the results from inverting the cdf.

It has to do with the (1-alpha)/alpha exponent not being applied to the trignometric functions and also maybe with the generation of the exponential random variable. The formula here is definitely not correct.

I have made a few tests on my machine with some corrections and they seem to work, so I will submit a fix in the near future and we can evaluate the results.

I created this issue to raise awareness, as I plan to submit a fix to this in the next few days/weeks.