Covid-19 modeling

I have officially thrown my hat into the ring and joined the throngs of would-be Covid-19 modelers to try to estimate (I deliberately do not use predict) the progression of the pandemic. I will pull rank and declare that I kind of do this type of thing for a living. What I (and my colleagues whom I have conscripted) are trying to do is to estimate the fraction of the population that has the SARS-CoV-2 virus but has not been identified as such. This was the goal of the Lourenco et al. paper I wrote about previously that pulled me into this pit of despair. I argued that fitting to deaths alone, which is what they do, is insufficient for constraining the model and thus it has no predictive ability. What I’m doing now is seeing whether it is possible to do the job if you fit not only deaths but also the number of cases reported and cases recovered. You then have a latent variable model where the observed variables are cases, cases that die, and cases that recover, and the latent variables are the infected that are not identified and the susceptible population. Our plan is to test a wide range of models with various degrees of detail and complexity and use Bayesian Model Comparison to see which does a better job. We will apply the analysis to global data. We’re getting numbers now but I’m not sure I trust them yet so I’ll keep computing for a few more days. The full goal is to quantify the uncertainty in a principled way.

2020-04-06: edited for purging typos

9 thoughts on “Covid-19 modeling

  1. Carson, what information criterion, cross-validation, or Bayesian out-of-sample predictive metrics are you going to use? FYI, I’ve found Aki Vehtari, Jonah Gabry, and Gelman’s improved version of log pseudomarginal likelihood which they call LOO to be a solid one with a pretty easy-to-use package in R:

    Their papers convinced me that LOO/WAIC are better to use than AIC and DIC.


  2. I too am excited to hear your results. Please don’t forget to include your less educated, less scientific layperson friends. Inquiring minds want to know…I WANT TO KNOW!


  3. Hi Carson, good to hear from you, and if anyone can bring rigor to this field, it is you. So far I am under-impressed with epidemic modeling. When I hear 100k < D < 900k, I see 500k +- 400k, so zero is just barely more than a one-sigma event! Best…..John


  4. I add my support to the sentiment raised by Reggie. We want to know (and some of us even want your work to get picked up by the popular press).

    Because I am an experimentalist, I will personally be interested in learning which data or combination of data constrains a particular result


  5. @Carson Did you guys write your LOO code yourselves or use an existing package? If you wrote yourselves, will go find on Github. I’m trying to thoroughly understand importance sampling.


Leave a Reply

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

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

Facebook photo

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

Connecting to %s