COVID-19 Coronavirus: Disastrous new forecast for Italy, 50k+. S Korea steady. China good.
Hi there. On Sunday I re-ran the forecasts for China, South Korea and Italy using the latest figures from WHO Report #49.
Here are the results for the potential end of various epidemics, subject to current conditions:
China cumulative cases: 81,466 by end of April (change <1/day by mid-April?)
China cumulative deaths: 3,395 by mid-May (change <1/day by mid-April?)
South Korea cumulative cases: 9,938 by 5 May (change <1/day by mid-April?)
2 South Korea
The South Korea forecast has improved due to a couple of data points slightly under the trend. Using the latest point as the start of the forecast gives South Korea at 9,938 cases for the potential end of the main part of the epidemic. This is where the trend is leading. It's not wise to obsess about precise numbers, because the trends can change - but the S Korea forecast has usually been in the region of 10,000 to 12,000 cases.
3. Italy
I wrote that is was probably early to attempt forecasts for Italy because it was very early days in the establishment of a trend and the trend might not be reliable. Very unfortunately, there has been a slight uptrend in the day over day % rate of change that has moved the trendline upwards. The slope of the trend line is key as it strings the epidemic out for much longer and the number of cases at the end is much increased. My two previous forecasts for Italy, extremely tentative, based on the then trend, were for 16,000 and 28,000 cases. Just a handful of days of worse than expected data and the end point now stands at 51,332 cases!
This allows for the possibility that, if data continues to surprise to the upside, even for a little while, Italy could end up with more cases than China.
Here is the new graph for Italy:
Compare this to the old forecast for Italy at 16,000 from around 4 March!
South Korea latest chart. The slight dog leg in the chart around 7,000 was caused by my adjusting the forecast starting date to 08 March:
China cumulative cases:
I was amazed that the data point for 8 March was right on the forecast line. This forecast starts with the data point on 23 February, 14 days previously. I think this shows that this forecasting tool is worthwhile:
A brief look at why Italy is proving disastrous:
Trend for Italy on 4 March: looking promising. Forecast was very tentative because line slope is shallow and day/day increase is still 20%+. This graph gave the first possible forecast of 16,000 for Italy:
Trend for Italy on 8 March: Things have changed. Day/Day % increase has actually trended back up, not completely out of the range of the other data points, which are fairly volatile but this makes a huge change to the numbers. S Korea and China have NOT behaved like this. They have so far been remarkably consistent. China for 1 months and S Korea for over a week.
This graph below gives the 51,000 figure for Italy:
Here are the results for the potential end of various epidemics, subject to current conditions:
China cumulative cases: 81,466 by end of April (change <1/day by mid-April?)
South Korea cumulative cases: 9,938 by 5 May (change <1/day by mid-April?)
Italy cumulative cases: 51,332 (!!) by 31 July (change <1/day by 28 June)
1. China
The China result surprised me because the daily rate of change has lowered the forecast considerably, implying fewer than about 1,000 more new cases. I find that a little difficult to believe but the maths says so. Of course, there is always the possibility of minor outbreaks that will make the epidemic linger for longer in China but the numbers are converging fast.2 South Korea
The South Korea forecast has improved due to a couple of data points slightly under the trend. Using the latest point as the start of the forecast gives South Korea at 9,938 cases for the potential end of the main part of the epidemic. This is where the trend is leading. It's not wise to obsess about precise numbers, because the trends can change - but the S Korea forecast has usually been in the region of 10,000 to 12,000 cases.
3. Italy
I wrote that is was probably early to attempt forecasts for Italy because it was very early days in the establishment of a trend and the trend might not be reliable. Very unfortunately, there has been a slight uptrend in the day over day % rate of change that has moved the trendline upwards. The slope of the trend line is key as it strings the epidemic out for much longer and the number of cases at the end is much increased. My two previous forecasts for Italy, extremely tentative, based on the then trend, were for 16,000 and 28,000 cases. Just a handful of days of worse than expected data and the end point now stands at 51,332 cases!
This allows for the possibility that, if data continues to surprise to the upside, even for a little while, Italy could end up with more cases than China.
Here is the new graph for Italy:
Compare this to the old forecast for Italy at 16,000 from around 4 March!
South Korea latest chart. The slight dog leg in the chart around 7,000 was caused by my adjusting the forecast starting date to 08 March:
China cumulative cases:
I was amazed that the data point for 8 March was right on the forecast line. This forecast starts with the data point on 23 February, 14 days previously. I think this shows that this forecasting tool is worthwhile:
A brief look at why Italy is proving disastrous:
Trend for Italy on 4 March: looking promising. Forecast was very tentative because line slope is shallow and day/day increase is still 20%+. This graph gave the first possible forecast of 16,000 for Italy:
Trend for Italy on 8 March: Things have changed. Day/Day % increase has actually trended back up, not completely out of the range of the other data points, which are fairly volatile but this makes a huge change to the numbers. S Korea and China have NOT behaved like this. They have so far been remarkably consistent. China for 1 months and S Korea for over a week.
This graph below gives the 51,000 figure for Italy:
Comments
Post a Comment