Copula models for astronomical distributions

Yesterday I read through this new arXival by an old friend from my ETH Zurich days, in which is presented a package (called LEO-py) for likelihood-based inference in the case of Gaussian copula models and linear regressions with missing data, censoring, or truncation.  I never quite understand the demand for astronomer-specific expositions and software on topics like this, since as soon as one understands what a hierarchical Bayesian model is and how to code one up in a standard statistical programming language like Stan or JAGS, then the world is your oyster.  (Indeed, back when we were at ETH, an errors-in-variables logistic regression model for predicting the barred galaxy fraction as a function of noisy-estimated stellar mass was one of my first forays into this field; of course it never saw the light of day because my supervisor at the time—she who shall not be named—was completely opposed to any statistical methods beyond ordinary linear regression!)  The key contribution here, to my mind, is rather the emphasis on copula models which are certainly under-utilised in the literature.  If this package helps popularise copulae (copulation?) that will be a very good contribution.

Note: Whenever I think of truncated astronomical data analysis problems I’m reminded of the example (described in JS Liu’s Monte Carlo Strategies in Scientific Computing) of a permutation test for doubly truncated (redshift, log-luminosity) data developed by Efron & Petrosian (1999).

This entry was posted in Uncategorized. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s