SEIR-model-Stockholm

Skattning av peakdag och antal infekterade i covid-19-utbrottet i Stockholms län februari-april 2020.

Rapport: https://www.folkhalsomyndigheten.se/smittskydd-beredskap/utbrott/aktuella-utbrott/covid-19/analys-och-prognoser/

This repo is a fork of Folkhälsomyndigheten’s repo containing the code for the SEIR model of the number of Covid-19 cases in Stockholm. I created this repo to understand their model better, find bugs or improvements to their code. I list my findings below.

Summary of changes

Bugs

In the original code, parameters are drawn from a distribution to calculate uncertainty estimates. The implicit reason why one can do that is that the optimization procedure is a maximum likelihood estimation, and such estimators have an asymptotic normal distribution. However, this distribution is multivariate normal, but the original code draws each parmeter independently from each other. The code in this repo fixes this.

Improvements

Parameters are transformed to each lie on the real line before being passed to optim. This fixes the non-convergence that the original code sometimes displayed.
Code has been parallelized via the furrr package where applicable, resulting in a pretty good speedup.
Uncertainty estimates are calculated for more quantities, such as the infectivity.

Cosmetic changes

Code now uses tidyverse packages to a much greater extent.
Functions are documented (work in progress).
Cleaner variable and function names.
Nicer plots via ggplot2.

Model outputs

Sensitivity analysis

Will explore model sensitivity to parameters gamma, eta and p_symp here.

3622e30b12c46853c5fd0e4e42d2670975e0ae6e