4 ms·
the (in)famous Imperial model is not much different in nature. It's an agent-based and capable of making nice pictures like this. https://github.com/mrc-ide/co
by buboard 6y ago
the (in)famous Imperial model is not much different in nature. It's an agent-based and capable of making nice pictures like this.
https://github.com/mrc-ide/covid-sim/ https://github.com/mrc-ide/covid-sim/
However i have not been able to find a case where their model could reproduce the epidemiology of a previous epidemic like influenza.
- andyljones 6y agoFunnily enough, this entire project sprung out of my trying to replicate the Imperial model a month ago. I found the slow part was the community transmission kernel, and was surprised to find no epidemiological literature on applying fast multipole methods to the problem. I was even _more_ surprised to find there was no easy-to-use black-box fast multipole method implementation out there. So I wrote one! It felt like a nice balance between 'doing something useful' and 'stepping on epidemiologists' toes'. Incidentally, I believe the Imperial model solves things by dividing the region up into cells of ~half a km across and assuming everyone in the cell is at the center. I didn't know this when I started the project, but kinda assumed it'd do something like that when I couldn't find any work on FMM in epidemiology.
- buboard 6y agothanks for doing this. YEs, the imperial model seems to be "Simcity for pandemics", but i m not sure why it was considered reliable since it wasnt ever calibrated to match a real world epidemic
- amluto 6y agoDid you tell the authors about your algorithm?
- andyljones 6y agoI sent some emails and didn't hear anything back. I'm hoping this post'll get organic attention for it, and I intend on sending more emails and - now you've prompted me! - opening an issue on the tracker too.