Process

Overview

The analysis extends principal component analysis to use a Bayesian framework.

Our starting point is therefore to produce a classical principal component analysis without extension.

We then repeat the analysis incorporating Bayesian priors, and compare results between the two.

Classical PCA

In this analysis we produce 2 classical principal component analysis models:

• Model 1A: for a data set that ends 2015, 7 years after the financial crisis of 2008
• Model 2A: for a data-set that is on the most uptodate dataset available from the Bank of England

Bayesian PCA

We repeat exten each of these models to incorporate a Bayesian framework.

• Model 1B: showing how incorporating appropriate prior could prevent overestimation*
• Model 2B: showing how can adjust current model for the future*

Determining Priors from Prediction Errors

⚠️ ** form of prior **

Future Looking Priors

⚠️ which insight