/
metadata-USACE-ERDC_SEIR.txt
40 lines (37 loc) · 2.57 KB
/
metadata-USACE-ERDC_SEIR.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
team_name: US Army Engineer Research and Development Center
model_name: ERDC_SEIR
model_abbr: USACE-ERDC_SEIR
model_contributors: Michael L. Mayo <Michael.L.Mayo@erdc.dren.mil>, Michael A. Rowland
<Michael.A.Rowland@erdc.dren.mil>, Matthew D. Parno <Matthew.D.Parno@erdc.dren.mil>,
Ian D. Detwiller <Ian.D.Dettwiller@erdc.dren.mil>, Matthew W. Farthing <Matthew.W.Farthing@erdc.dren.mil>,
William P. England <William.P.England@erdc.dren.mil>, Glover E. George <glover.e.george@erdc.dren.mil>
website_url: https://github.com/erdc-cv19/seir-model
license: other
team_model_designation: primary
ensemble_of_hub_models: false
methods: The ERDC SEIR model makes predictions of several variables (e.g., reported
new/cumulative cases per day, etc.). Model parameters are estimated using historical
data using Bayesian inference.
team_funding: US Army Geospatial Task Force
data_inputs: USA FACTS, JHU COVID-19 Git Repo
methods_long: 'The ERDC SEIR model is a process-based model that mathematically describes
the virus dynamics in a population center (e.g., state, CBSA) using assumptions
that are common in compartmental models: (i) modeled populations are large enough
that fluctuations in the disease states grow slower than averages (i.e., coefficient
of variation < 1) (ii) recovered individuals are neither infectious nor become
susceptible to further infection.
The model is similar to classic SEIR models but with additional compartments for unreported
infections and isolated individuals. Model parameters describing reporting rates
and transmission rates are allowed to vary in time over over the historical record
as social and policy factors change. Forecasts into the future, however, are made
under the assumption that the parameters are held constant in time.
The model parameters are estimated from observations of the cumulative number of cases
using a Bayesian approach. Information from subject matter experts (SMEs) is used
to develop a prior probability distribution over the model parameters. This then
combined with a statistical model of the model-data mismatch to produce a posterior probability
distribution over the model parameters that incorporates both the cumulative observations
and subject matter expertise. The parameters that maximize the posterior probability
density are used to make forecasts into the future. The effective reproduction number
is computed using the next-generation matrix method.
Deaths are estimated from this model through a procedure which leverages state-level data
on recoveries and fatalities.'