I have a bunch of correlation coefficients, which I assume are samples of a true underlying correlation, which I’d like to estimate using Bayesian inference. The sampling distribution has an analytical solution but is composed of non-standard functions; consequently I’m having trouble converting the…

Fitting a distribution with custom functions

twiecki August 3, 2017, 11:53am 5

If there is a gradient, otherwise you have to use Slice or Metropolis sampling.

Topic		Replies	Views
Pymc3 likelihood math with non-theano function Questions theano	4	1088	September 25, 2017
Defining custom multivariate Distribution with gradients Questions theano	5	1605	August 24, 2018
Can't vectorise custom distribution Questions theano	2	748	May 24, 2018
Passing pymc3 distribution variable to theano function Questions	3	1648	September 23, 2019
Integrating new distribution (wrapped c++ functions) - Problems with broadcasting and shapes Questions theano	3	715	July 8, 2019