Econometrics / Machine Learning Interface

Check out this exciting new seminar series. Great initiative!

AMLEDS Seminar (Applied Machine Learning, Economics, and Data Science)

Serena Ng (Columbia University) will give the inaugural seminar. The discussion moderator is Michael McMahon (Oxford University).

Topic: "Methods for Analyzing Data with Missing Values and from New Sources"

13:00 EST time / 18:00 GMT time / 19:00 CET time, November 6, 2020. 

To register, and to learn more about AMLEDS, go to https://sites.google.com/view/amleds/home.

Factor Loadings and Network Connectedness

I have long been interested in using time-varying latent factor loadings for time-varying connectedness measurement. Patrick Gagliardini made great progress on time-varying loadings in his October 19 SoFiE talk, discussed by Seth Pruitt (recording etc, here).  

I want to relate time-varying latent factor loadings to the Diebold-Yilmaz connectedness measurement framework. Kelly et al. (2019) helps, showing how to empirically assess the correlation, if any, between time-varying factor loadings and time-varying DY network connectedness, by allowing loadings to depend on covariates like connectedness.

But what I really want is a "Rosetta Stone" giving a 1-1 translation between Patrick's "time-varying factor loadings" world and the DY "time varying network centrality" world. That's almost surely wishful thinking in general, but maybe under some (stringent) conditions?

PREDOC

Check out www.predoc.org.  It just launched.  Moving forward I think this will be a key clearinghouse for information for diverse undergrads seeking an economics "pre-doc" before potentially heading on for the Ph.D. Current members here, and growing daily, as are available predoc positions.