Epidemiologists should consider GLM & IMR for better Covid death predictions and overcome testing bias

Interested in publishing a one-time post on R-bloggers.com? Press here to learn how.
Social science research network (SSRN) reviewer has posted our “A Novel Solution to Biased Data in Covid-19 Incidence Studies” written by myself (H D Vinod) and K Theiss.  It uses a new two-equation generalized linear model (glm) with Poisson link for forecasting Covid-19 deaths in the US and individual states. The first equation explicitly models the probability of being tested. It helps build an estimate of the bias using inverse mills ratio (IMR) Detailed paper is available for free download at:

ssrn.com/abstract=3637682 It uses open-source R software.  A supplement to the paper entitled “State-by-state Forecasts of Covid-19 Deaths” is available for free download at http://www.fordham.edu/economics/vinod/WeeklyCovid.pdf Our bias-corrected out-of-sample death forecasts for future weeks should help state governors and mayors decide on limits to opening local economies.  One can compare the performance of democratic versus republican governors using the information tabulated.

Published by

Hrihikesh Vinod

Ph. D (Harvard) Economics, Professor of Economics, Fordham University. Author of two R packages meboot and generalCorr

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.