AI4D Predict the Global Spread of COVID-19
Helping Global
$5 000 USD
Challenge completed over 5 years ago
Forecast
934 joined
51 active
Start
Mar 13, 20
Close
Apr 19, 20
Reveal
Jun 08, 20
China's Reviewed Numbers
Help · 21 Apr 2020, 12:04 · 5
On the 17th of April, China reviewed its numbers and it jumped on the JHU website from 3346 to 4636 in one day. This review didn't capture the exact number of deaths on that day, it was just to show China's revision. This might greatly distort models that are supposed to be based on real trends. Apparently, 1290 people didn't die in one day in China. For a long time before that, the death numbers had stabilised.
What does Zindi propose should be done about this?
Dear Alchemi,
Thank you for bringing this to our attention.
We will be taking the JHU numbers as they are. We are not experts and we cannot make a definitive call as to the accuracy of any numbers that JHU publish.
Everyone built their models on the same data, so their scores will all be subject to the same randomness of dirty data.
Thanks for the response.
Including China will significantly affect the results. Wish we could exclude China on the premise of "fixed" and unrealiable numbers. My thoughts...
Just checked China's number again. It has been fixed at 4636 for the past 1 week. Including such a country will have undesirable effect on MAE. @Zindi, we would appreciate if China's number is removed from the scoring. Your consideration will be appreciated.
Promoting open access to health data can allow independent researchers to analyze trends more effectively and hold @monkey mart governments accountable.