Slovakia: Small number of cases = bad fit?
gergelytakacs opened this issue · 6 comments
Hello there @ECheynet ! Thank you for sharing your wonderful repo.
I've been trying to fit your model to the data for Slovakia, where, thanks to the fast and aggressive action, there is a little number of confirmed cases. Right now, there are <500 confirmed infections, 0 deaths, but statistics only show 8 recoveries.
I had no luck trying to find a reasonable fit, I was even spending some time to play around with the initial guesses. Do you think the low number of cases and no data in the "death" set could be the reason for this bad fit? Do you have any suggestions on how to improve this?
I've packed my data and script in a zip too:
SK-COVID-19.ZIP
Here is the fit quality (with data) I get:
Recoveries are a good fit, but infections are not even close.
Hi,
The fit should be much better than that.
This is what I got when using Example2.mlx and I replaced South Korea with Slovakia:
Maybe retry to run the example2 with the most recent version of the code?
I think I know where the problem comes from. If you do not have access to the function "contains", it uses an alternative formulation, which contains a bug. I will update the code in the next few minutes.
Thank you for having a look at this. Your fit actually looks awesome! I was struggling to get a reasonable fit with my own SIR/SEIR etc. models before, but this looks very good indeed! Excited to see the update, be sure to let me know when its out.
It's out since last night
Awesome. Will pull and test, will get back here.
@ECheynet Works as a charm. Have used Example3.mlx for now. Will try to find why my own code produces bad fit (by own I mean the m - file attached to this issue) but that's possibly another question. Thanks for the help!