search

Petrichor12 / Coronavirus

Analysis on the novel coronavirus data from worldometers.
2 stars 0 forks source link

Plot curve fit projections and actual cases #6

Open gokeplerride opened 4 years ago

gokeplerride commented 4 years ago

@Petrichor12 I created a data frame that has exponential curve fits as well as the actuals starting at 100 cases. It would be cool to plot those and see how things could develop. My plotting skills are pretty basic though...

Petrichor12 commented 4 years ago

Cool idea, will work on it during home office this week lol

Petrichor12 commented 4 years ago

Another idea, we should work on a few different mathematical models, for example including a paramater such as 'introduced lockdown on x day', and 14 days after that there is an inflection point and cases start decreasing. We could base it on China's case data.

gokeplerride commented 4 years ago

Agree although it might be difficult due to the low number of observations... not really sure how that could be done

Dan-OKeeffe commented 4 years ago

hmm yeah - i wonder how you can also account for the amount of testing done (it seems to massively affect the visible growth rate - see the US at the moment :S)

Petrichor12 commented 4 years ago

There is some testing data on worldometers but it's not updated enough - I agree it would be cool though

Petrichor12 commented 4 years ago

https://en.wikipedia.org/wiki/COVID-19_testing

Here is a lot more testing stats, but still not all are updated.

Petrichor12 commented 4 years ago

https://neherlab.org/covid19/

An idea based on this website from @Dan-OKeeffe - check out the mitigation section. We could inspect the source code, find out what model they use for the mitigation, and use that to make our own predictions. Could also incorporate more of the parameters into the model.

Petrichor12 commented 4 years ago

We could use/analyse growth rate to do more analysis and predictions. To smooth it out for example case growth rate over the last 7 days.

gokeplerride commented 4 years ago

I suppose that might work. It would actually be interesting to plot growth rate with a timeline reset around 200 cases

Petrichor12 commented 4 years ago

Definitely, because the data is too noisy before that. Could even start it at 1000 cases with a rolling average to smoothen it out.