bgoodri

Upon further review, can this method generate square orthogonal matrices with negative determinants? I can't seem to generate one when doing this in R: ``` rstan::expose_stan_functions("https://raw.githubusercontent.com/pourzanj/TfRotationPca/master/Stan/UniformStiefel.stan") sign_det

Taking the square root is where most of the precision gained from using LDLT gets lost. There should be a way to differentiate with respect to the diagonal elements of...

I think all of these hypergeometric functions should (eventually) call the ones in Boost, which have been included in the last several versions. https://www.boost.org/doc/libs/1_79_0/libs/math/doc/html/math_toolkit/hypergeometric.html It is hard to do well...

Is 5e-309 really a problem? On Fri, Mar 25, 2022 at 12:44 PM Andrew Johnson ***@***.***> wrote: > It looks like the current 3.4 release has a bug with handling...

I think temporarily changing the unit test would be preferable to disabling SIMD for something that is correct for all numerical intents and purposes. On Fri, Mar 25, 2022 at...

No objection

That is more or less what rstan has done with BH when stuff like this happens. On Wed, Jan 12, 2022 at 11:34 AM Andrew Johnson ***@***.***> wrote: > Description...

We don't have any of that stuff at the moment because everything in the latest BH works as far as we know of. But over the years, we sometimes put...

Paul Godfrey, who was the guy that really came up with this approximation, said (with reference to the Gamma function, rather than the log-Gamma function) http://my.fit.edu/%7Egabdo/gamma.txt > The best performance...

What claims are unsubstantiated? He got that precision for the Gamma function. You linked to the Boost doc for the digamma function. For the log-Gamma function with fixed precision Boost...