Let’s say that I have a timeseries input with 1000 timesteps. I then make an experiments for 100 steps in which I have control, and treatment group. So I have 1200 datapoints, but the last 100 timesteps overlap each other. If I model using additive model, how should I update the model parameters? I …

I would imagine a model with 3 timeseries, the 1x1000 initial and 2x100 experiments? Your last model seems to be double counting the 1x1000 observations, which sounds wrong.

How to update PyMC model with counterfactuals?

Questions v5

drbenvincent April 26, 2023, 6:05am 6

In case it’s useful, there are a number of example notebooks with counterfactual here PyMC Example Gallery — PyMC example gallery

Topic		Replies	Views
Multi-output gaussian processes Questions	13	5511	October 22, 2017
Sampling semantics of multiple observed variables Questions	5	2594	April 25, 2019
Multiple (uncertain) function observations of the same Gaussian process Questions	15	6568	October 16, 2017
How to get "on-line" prediction estimates for data over time? Questions	6	803	May 19, 2020
Predicting from a model with multiple observed variables? Questions	4	4061	October 9, 2018

How to update PyMC model with counterfactuals?

Related topics