-
Peer review of Ferguson's code behind his model currently going on in Github is savage.
Any of the savagers come up with a better model yet?
-
The savaging is from parachuted in libertarian weirdo Hacker News types with zero experience of coding anything similar or any understanding in of what it's meant to do. Certainly not peers.
Almost all of the complaints are cosmetic "I wouldn't do it like that" stuff. Nothing that invalidates the model.
They recommend you run it multiple times until you get the results you hoped for..
It simulates random spread, so by design it produces different outputs each time and you're meant to average them as needed.
The bug you link to is related to loading and saving state where it should produce the same output for the same inputs, but the random number generators weren't being correctly reset.
You are reading a single comment by @user62119 and its replies.
Click here to read the full conversation.
Brommers
grams
@user62119
Peer review of Ferguson's code behind his model currently going on in Github is savage.
Turns out what was 'open sourced' isnt the original code but a heavily refactored version by Microsoft and Github removing Fortran sections (took some of the best coders in the world a month)
https://github.com/mrc-ide/covid-sim/issues/144#
But even after this work, its still buggy as fuck.
The official docs say the output isnt reliable.
They recommend you run it multiple times until you get the results you hoped for..
https://github.com/mrc-ide/covid-sim/issues/116
Team from Edinburgh uni showing how same entry can generate a different modeled outcome
