Catalog of public COVID models

This is a catalog of various public models for COVID-19. This catalog is especially concerned with models which have good coverage for California and its counties.

Reproduction Number (Nowcasts)

These are models of \(R_t\) the time-varying reproduction number, which is the average number of people infected by an individual. An \(R_t>1\) indicates the disease is spreading, whereas \(R_t<1\) indicates the disease is dying out.

Model Group Website Source Code Methods Raw Data
Covid Act Github methods json, api
covidestimcovidestim.orgpackagemedRxivcsv Github methods csv dir
Harvard Lim Labcovid19-analysis.orgnomethodscsv dir
Imperial College Github methods csv
Kevin Systrom et Githubnocsv

Some other \(R_t\) resources

  • CRAN R0 package: A package for R which uses number of cases and a few other epidemiological parameters to estimate R historically.

  • ICL ensemble: A collection of four models to estimate the growth rate.

  • RECON (R Epidemics Consortium): This group has created a number of packages including EpiEstim to estimate R from data.

Forecasts of Epidemological Outcomes

These are forecasts of deaths, infections, reported cases, hospitalizations an other related epidemiological outcomes. These are mostly compartmental models, such as SIR, SIER and variations. Sometimes this type of model is called a “mechanistic model” because it simulates a mechanism of disease transmission. A few of the models below are more akin to statistical extrapolation, without an underlying model of disease dynamics.

Model Group Website Geographic
Open Source Methods Raw Data
Caltech CS156caltech.eduCountynonono
Shaman group
Covid Act CountyGithub methods json, api
Facebook AI Researchai.facebook.comCountynomethodscsv
GLEAM projectgleamproject.orgStatenomethodsjson
IHMEcovid19.healthdata.orgStateCurve fit
SEIIR pipeline
SEIIR core
Imperial College
London StateGithub methods csv
LEMMAlocalepi.github.ioCA countyGithubGithub of pdf & xls
MIT DELPHIcovidanalytics.ioStateGithubmethodssee bottom of
projections page
RAND Corprand.orgStatenomethodsno
U Texasutexas.eduState,MSA'snomethods, FAQGithub of csv's
U Mass MechBayesGithub.comStateGithubmethodscsv on Reichlab
UC Berkeley
Yu Group
UCLA MLcovid19.uclaml.orgCA counties, LAnomedRxivjson
UCSDucsdcoidreadi.comCA countiesnosee “About This App”csv
Youyang Gucovid19-projections.comSome
SEIR simulator
(not full model)

Reichlab Ensemble

The Reichlab at University of Massacheusettes Amherst has created This is a repository where many forecasters to submit their forecasts in a standard format, to faciliate easy comparison. The forecasts are near-term estimates of deaths at the state and national level. Some forecasters submit quantiles as well. Other epidemiological measures such as cases or hospitalizations are not included in the Reichlab ensemble. Previous forecasts can be found at

Youyang Gu periodically evaluates the historical performance of the various forecasts.

Agent-based models

Rather than modeling the dynamics of various groups (susceptible / infectious / recovered), agent-based models simulate specific individuals. This allows for very detailed models of particular interactions. Below are some software frameworks for doing agent-based modeling.

Model Group Website Open Source
Oxford BDM link OpenABM
The Institute for Disease Modeling (IDM)) CovaSim