Stochastic volatility predictions?

#1

Hi friends,

I have revised through the whole of chapter 10 again and again but I can’t seem to figure out how to get out-of-sample predictions for volatility using NUTS.

I’m trying to sample from 2000-2019 and then predict 2019-2020, could anyone help with this?

#2

The notebook pymc3_workflow has a prediction example, perhaps this could help?

#3

Hey Stefan, thanks for the reply. In pymc3_workflow, there was a predictive variable: the yield curve. However, for stochastic volatility, it seems we are only predicting based on sampling with the NUTS sampler. I’m not sure if we can forecast purely based on historical values for volatility?