I am working through McElreath’s Rethinking and have hit a pretty early wall with some basic PyMC. Specifically I want to use the quadratic approximation on a Bernoulli trial experiment. When I calculate the MAP I get a reasonable estimate for p, but when I try to estimate standard deviation via inv…

This appears to be related to the change of variables PyMC uses internally when sampling. You should get the right answer from this model: with pm.Model() as normal_approximation: p = pm.Uniform('p', 0, 1, default_transform=None) w = pm.Binomial('w', n=len(data), p=p, observed=data.sum()) T…

Discrepency Between Hessian Estimation of Std. Dev. and Sampled Estimation

Questions v5

ricardoV94 April 14, 2025, 2:08pm 7

The hessian can be computed on unconstrained or constrained space and those are generally different

Topic		Replies	Views
Find_hessian version differences v5 bug	14	751	November 7, 2022
Using find_MAP and hessian to find frequentist standard errors, t-stats and p-values Questions	2	1472	March 31, 2020
Problems with finding normal approximation to binomial v5	3	489	December 22, 2022
NaNs in Hessian when using NormApprox Questions	5	810	May 3, 2019
How to obtain the variance estimations in Linear Regression models Questions	5	889	June 26, 2017

Discrepency Between Hessian Estimation of Std. Dev. and Sampled Estimation

Related topics