• Andrea De Vita
• Enrico Lupi

---

Bayesian Blocks Algorithm

Final Project for Advanced Statistics for Physics Analysis course, Physics of Data degree at University of Padova.

Assignment:

Bayesian Blocks: a dynamic algorithm for histogram representation.

It is a non-parametric representation of data derived with a bayesian statistical procedure. It has been invented by D. Scargle ¹² and applied in the context of astronomical time series analysis. A similar technique available on the market is the kernel density estimation (KDE). As described in ³, it allows to discover local struture in background data, exploiting the full information brought by the data. The main idea is based on segmentation of the data interval into variable-sized blocks, each containing consecutive data satisfying some well defined criteria.

Write an algorithm and then test its performances with different sets of data.
Try to think other possible application of the method.

Footnotes

1. J. D. Scargle et al., Astrophys. J. 764 (2013) 167 ↩

2. J. D. Scargle et al., Astrophys. J. 504 (1998) 405 ↩

3. B. Pollack et al., arXiv:1708.008 10 ↩