I’m doing a simulated data study in which I am seeking to explore whether my regression model is able to represent a (known) data-generating model with limited data. I first generate fake data, then fit the fake data using HMC (NUTS). For the sake of simplicity, let’s assume that the true model is …

You could certainly do this without looping. You just need to create J-length arrays of stochastic nodes using the shape= kwarg rather than creating one node per stochastic per model, then map them to the correct data in y using appropriately designed design matrices. Personally I would reshape you…

Fit many regressions simultaneously

Questions

sammosummo November 3, 2017, 6:40pm 4

Its a bit weird that you name both the stochastic node representing the intercept and intercept plus effect of the covariate mu, but I don’t think you will get this to run much faster.

Topic		Replies	Views
Multiple regression caveats Questions	6	856	July 23, 2018
Fitting data with multiple attributes Questions	3	647	January 17, 2018
Help translating from JAGS to PyMC3 Questions	4	1225	August 21, 2019
Managing indices/data shape in a multi-level linear regression for incremental model updating Questions	0	381	September 10, 2020
MRPyMC3 — Multilevel Regression and Poststratification with PyMC3 Sharing	0	1024	July 10, 2017

Fit many regressions simultaneously

Related topics