Burbles about COVID data
This is a little musing about a couple of items related to COVID 19. The first is a reaction to an article on the ABC site headed "something like "You only need one number to monitor COVID". The proposed indicator is the change in the average rate of new infections for the previous 5 days compared to the same data for the previous day. If the value is below 1 it means the rate of new infections is dropping.
It isn't a bad indicator but does have a flaw. Here is an image of the Australian data for the period starting on 19 April. (The ABC show a much longer time series with the data below for quite a while.)
The value of the indicator can first be calculated on 24 April and shows a value well below 1 as the calculation of the 5 day sum excludes a value of 41 and includes a value of 14. However the next day 13 drops off and is replaced by 20 so the value goes well above 1 but dropping the next day as 26 is replaced by 16. I show the number of new cases for the 27th in grey as I have, for the sake of illustration, used the average of the previous 5 days. That value of 15 is well above the (astonishingly low) value of 4 cases dropped off on the 22nd so up goes the proportion again.
Where I get to is that the 'magic number' concept doesn't work well when the number of new cases is quite small. The ABC cautions that to be of concern the number must be above 1 for several consecutive days, but I suspect that is getting close to tossing a coin and if the answer isn't correct going for the best of 3, 5, 7 ..... until the desired outcome is achieved.
The second issue is the launch of the COVIDsafe app. That got to 1 million downloads in the first 12 hours. Which is pretty good. However the Government has said it needs 40% coverage - 10 million downloads - to be effective.
My first thought is that the app was touted as being good because it was thought to have worked in Singapore where only 20% of people downloaded it. (It now seems that Singapore hasn't done too well as the number of infections there has steadily risen. According to Crikey that is because the Government didn't pay attention to infections in migrant worker hostels, and not a flaw with the app per se.)
Greg Hunt, the Minister for Health - who has been doing a far better job than he did as the Minister forAppeasing Adani Environment - was quoted on the ABC as saying "We got the first million within five hours. We had been hoping, our best hope, was we might get to 1 million in five days."
Hmmm. When I worked in mail-collection surveys in the ABS we found we got a very high response rate early on and then it gradually dropped until we sent out out a reminder and it perked up again. So I played with some numbers simulating various declines in response. I stress that these are mind games and aren't meant to prophesy what I expect to happen. But I think they show some indicative possibilities.
For ease of demonstration I used steady drops: in my first case I assumed the response on day 2 was half that of day 1, second case a 20% drop and thirdly a 10% drop. The results in terms of cumulative number of downloads are shown in this graph.
If the results drop by 50% a day they will finish up with 2 million downloads and with the 20% decline it will die at 3 million. Only if you assume a drop off as low as 10% do you hit the 10 million mark and that would be after 26 days.
That may not be an issue, but the Government has tied its loosening of restrictions to inter alia getting the app working. Will they hold on to the current level of restrictions for a further 4 weeks, especially if the number of new cases stays around 10 a day?? I don't think so.
If I was the Government I would be trying to come up with some way of getting the download rate increasing. Since IMHO the big issue is that the public do not trust politicians in general and Messrs Scumo, Adolf Kipfler and Greg Hunt in particular I reckon that is a big ask.
Since I wrote that a week ago the app was downloaded another million times of the next day and has inched to 4 million after 7 days, but seems to have stalled. The news has been unrelentingly pathetic in terms of the Government building trust (not):
In contrast ScuMo is making noises about if you don't download it the restrictions will stay. But it isn't compulsory, no sir, trust me, I'm from the Shire,
It isn't a bad indicator but does have a flaw. Here is an image of the Australian data for the period starting on 19 April. (The ABC show a much longer time series with the data below for quite a while.)
Where I get to is that the 'magic number' concept doesn't work well when the number of new cases is quite small. The ABC cautions that to be of concern the number must be above 1 for several consecutive days, but I suspect that is getting close to tossing a coin and if the answer isn't correct going for the best of 3, 5, 7 ..... until the desired outcome is achieved.
The second issue is the launch of the COVIDsafe app. That got to 1 million downloads in the first 12 hours. Which is pretty good. However the Government has said it needs 40% coverage - 10 million downloads - to be effective.
My first thought is that the app was touted as being good because it was thought to have worked in Singapore where only 20% of people downloaded it. (It now seems that Singapore hasn't done too well as the number of infections there has steadily risen. According to Crikey that is because the Government didn't pay attention to infections in migrant worker hostels, and not a flaw with the app per se.)
Greg Hunt, the Minister for Health - who has been doing a far better job than he did as the Minister for
Hmmm. When I worked in mail-collection surveys in the ABS we found we got a very high response rate early on and then it gradually dropped until we sent out out a reminder and it perked up again. So I played with some numbers simulating various declines in response. I stress that these are mind games and aren't meant to prophesy what I expect to happen. But I think they show some indicative possibilities.
For ease of demonstration I used steady drops: in my first case I assumed the response on day 2 was half that of day 1, second case a 20% drop and thirdly a 10% drop. The results in terms of cumulative number of downloads are shown in this graph.
That may not be an issue, but the Government has tied its loosening of restrictions to inter alia getting the app working. Will they hold on to the current level of restrictions for a further 4 weeks, especially if the number of new cases stays around 10 a day?? I don't think so.
If I was the Government I would be trying to come up with some way of getting the download rate increasing. Since IMHO the big issue is that the public do not trust politicians in general and Messrs Scumo, Adolf Kipfler and Greg Hunt in particular I reckon that is a big ask.
Since I wrote that a week ago the app was downloaded another million times of the next day and has inched to 4 million after 7 days, but seems to have stalled. The news has been unrelentingly pathetic in terms of the Government building trust (not):
- while the app has been downloaded and presumably activated the State Health people are not yet in a position to accept the data;
- the Government hasn't passed the legislation containing the privacy protections (in fact no-one has seen it yet);
- the source code of the app hasn't been shown to anyone; and
- it appears the app doesn't work well on iPhones (presumably the goose in DSD who wrote it is an Android user!
In contrast ScuMo is making noises about if you don't download it the restrictions will stay. But it isn't compulsory, no sir, trust me, I'm from the Shire,
Comments