Bold forecast for South Korea: 10500 cases on current trend if last 2 days decline continue BUT THEY DIDN'T so revised to 14000.
Bold forecast for South Korea: 10609 cases on current trend if last 2 days decline continues:
This is very tentative statistically right now but it's on the record.
UPDATE # 1
However, today's point came in higher than expected. There were 5186 cumulative cases according to the Chinese DXY website, a much larger jump than each of the past two days. This gives a point above the trend line and increased the final forecast case count to 14,000+. In fact, to 14,113.
This increase comes from just small changes in the coefficients due to adding an extra data point. ...
Yesterday's chart gave a more optimistic forecast. Right now, there is a high level of uncertainty because only 12 or 13 data points. ...
However, with more data, if the % day over day trend continues to fall over time, the forecast accuracy should improve.
Right now, the forecast now stands at 14,000.
Using this technique for China has worked well owing to the lock-downs and the consistent monitoring there, minimising social contact and spread of infection. Even so, after the lock-downs started the case numbers increased from 1,000 to 80,000! My initial forecasts from 12 February were still fairly consistent with what has transpired since. I am wondering if that is because the lion's share of the epidemic was in one city. If multiple cities or provinces in a country have outbreaks at different times, forecasting will be next to impossible.
UPDATE # 2
The WHO figures have come in for end of day yesterday. I use these normally because you know the time-stamp for the data will be consistent. It gives 4840 cases for S Korea (the same as the number of square yards in an acre in old imperial measurements).
This moves the line a little again and gives a final forecast of 10,950 cases assuming they continue to do things right, which the really seem to be doing.:
This is very tentative statistically right now but it's on the record.
UPDATE # 1
However, today's point came in higher than expected. There were 5186 cumulative cases according to the Chinese DXY website, a much larger jump than each of the past two days. This gives a point above the trend line and increased the final forecast case count to 14,000+. In fact, to 14,113.
This increase comes from just small changes in the coefficients due to adding an extra data point. ...
Yesterday's chart gave a more optimistic forecast. Right now, there is a high level of uncertainty because only 12 or 13 data points. ...
However, with more data, if the % day over day trend continues to fall over time, the forecast accuracy should improve.
Right now, the forecast now stands at 14,000.
Using this technique for China has worked well owing to the lock-downs and the consistent monitoring there, minimising social contact and spread of infection. Even so, after the lock-downs started the case numbers increased from 1,000 to 80,000! My initial forecasts from 12 February were still fairly consistent with what has transpired since. I am wondering if that is because the lion's share of the epidemic was in one city. If multiple cities or provinces in a country have outbreaks at different times, forecasting will be next to impossible.
UPDATE # 2
The WHO figures have come in for end of day yesterday. I use these normally because you know the time-stamp for the data will be consistent. It gives 4840 cases for S Korea (the same as the number of square yards in an acre in old imperial measurements).
This moves the line a little again and gives a final forecast of 10,950 cases assuming they continue to do things right, which the really seem to be doing.:
Comments
Post a Comment