Prof Darrel Francis ☺ Mk CardioFellows Great Again
10 Oct,
33 tweets, 5 min read
This is a REALLY good question!
Let's assume all tests are perfectly accurate, for simplicity's sake.
And let's ignore the word "up to", again for simplicity's sake.
What does this statistic mean?
Let's assume all tests are perfectly accurate, for simplicity's sake.
And let's ignore the word "up to", again for simplicity's sake.
What does this statistic mean?
Does this mean that:
Well I am glad nobody has voted yet.
Because all the answers are wrong.
Let me highlight the relevant words in the question, for people who read it too carelessly.
Because all the answers are wrong.
Let me highlight the relevant words in the question, for people who read it too carelessly.
Most people in the world today
Roughly what proportion of people in your country, today, have symptoms of Covid?
So somewhere between 95 and 99% of the people in your country are asymptomatic.
In _screening_ programmes, what proportion of people are asymptomatic?
In _screening_ programmes, what proportion of people are asymptomatic?
First answerer is correct!
Screening is testing ASYMPTOMATIC people.
Screening is testing ASYMPTOMATIC people.
In a screening programme, what proportion of people who test positive, will be asymptomatic?
In a screening programme, what proportion of people who test NEGATIVE, will be asymptomatic?
Meanwhile, there is also testing services for people who HAVE SYMPTOMS.
What proportion of those patients are asymptomatic?
What proportion of those patients are asymptomatic?
In real life, both types of programmes exist.
If you present a report of SCREENING programmes, the proportion asymptomatic will be 100%.
If you present a report of DIAGNOSTIC TESTING FOR SYMPTOMATIC PATIENTS, the proportion asymptomatic will be 0%.
If you present a report of SCREENING programmes, the proportion asymptomatic will be 100%.
If you present a report of DIAGNOSTIC TESTING FOR SYMPTOMATIC PATIENTS, the proportion asymptomatic will be 0%.
If your report includes both groups of people, what will your proportion asymptomatic be?
What will the "proportion asymptomatic" depend on?
Most reports are from a mixed source, because they are just a service reporting the totality of their data.
Their statistic is correct, but no more interesting than the researcher's eye colour or favourite breakfast cereal.
No reason to quote it as evidence of anything!
8-)
Their statistic is correct, but no more interesting than the researcher's eye colour or favourite breakfast cereal.
No reason to quote it as evidence of anything!
8-)
I don't find it so incomprehensible.
I assume they are trying to work out how many people in a region have Covid currently, so they are testing a random sample of people, and mentally scaling it up?
Why use PCR? I don't know anything about virology. What else can they use?
I assume they are trying to work out how many people in a region have Covid currently, so they are testing a random sample of people, and mentally scaling it up?
Why use PCR? I don't know anything about virology. What else can they use?
I see your point @ChZoepfchen!
Yes, the most reliable way to estimate the number of infected people in a population would be to summon people at random for testing, and calculate the proportion (of those tested) positive.
Dividing by the population of the region is a bit mad.
Yes, the most reliable way to estimate the number of infected people in a population would be to summon people at random for testing, and calculate the proportion (of those tested) positive.
Dividing by the population of the region is a bit mad.
But I am sympathetic to non-random sampling.
Although random is best for reliable inference, it is VERY difficult to do.
Here, if the government phones people up and tells them they have been chosen for random testing, people would laugh at them.
We enjoy disobedience in UK.
Although random is best for reliable inference, it is VERY difficult to do.
Here, if the government phones people up and tells them they have been chosen for random testing, people would laugh at them.
We enjoy disobedience in UK.
Interesting question here.
https://twitter.com/jackwebster1990/status/1314859734051282944?s=20
Delicate issue because when the pandemic was new, I got caught for not knowing whether "case" meant any infected person (even if asymptomatic) or only symptomatic people.
Turns out "case" means "symptomatic infected" people.
Turns out "case" means "symptomatic infected" people.
But to rephrase your question as "Is it right to include those 5k people who have asymptomatic infections and make decisions on new restrictions based on this?"
I don't have an answer because I have no good idea how to decide on what restrictions to implement.
I don't have an answer because I have no good idea how to decide on what restrictions to implement.
A lot of people are stupid and/or contrary.
They do the opposite of what they are told, or make up their own reasons for doing things.
I mean, lots of people smoke, and some people even play the lottery.
They do the opposite of what they are told, or make up their own reasons for doing things.
I mean, lots of people smoke, and some people even play the lottery.
Their smoking or lottery-playing doesn't harm others. Indeed it helps the others, by providing a net outward transfer of wealth from people without long term thinking to people with long term thinking.
But for coronavirus, the actions of less analytical people are of crucial importance because they dictate the rate and extent of spread of the virus.
The government's instructions are largely for those people.
The government's instructions are largely for those people.
And they don't have to be based on particularly elaborate or precise science.
We just need to get R down under 1, i.e. we just need to settle the epidemic down, by reducing transmission below that threshold.
We just need to get R down under 1, i.e. we just need to settle the epidemic down, by reducing transmission below that threshold.
If people won't accept or adhere to advice based on that woolly feeling that is (nevertheless) correct, then I don't blame the government for pointing to (effectively) random numbers and saying they are the reason for restrictions.
If we can't handle the truth, we will be fed random factoids.
That's the job of the Government. To get people to settle the epidemic down, without putting troops on the streets and shooting people.
Pointing at random numbers is a small indiscretion, compared to that.
That's the job of the Government. To get people to settle the epidemic down, without putting troops on the streets and shooting people.
Pointing at random numbers is a small indiscretion, compared to that.
Great point:
I applaud ONS's effort, but it is not the police or the army. It can't force people to be tested.
As soon as you let people have a choice, you end up selecting for people who have a reason to be tested.
And now it is the INTELLIGENCE of people that you need to be afraid of.
As soon as you let people have a choice, you end up selecting for people who have a reason to be tested.
And now it is the INTELLIGENCE of people that you need to be afraid of.
Ironic, that!
When invited for a test, people with any common sense will be more likely to say yes if there is a reason for concern in them, for example,
- they felt slightly poorly
- a friend or family member had Covid
- their region of country is having an uptick
When invited for a test, people with any common sense will be more likely to say yes if there is a reason for concern in them, for example,
- they felt slightly poorly
- a friend or family member had Covid
- their region of country is having an uptick
As a result, the proportions we find will be upwards-biased.
But ONS will know that, and make it clear in their reports, I am sure. They are professional epidemiologists.
But ONS will know that, and make it clear in their reports, I am sure. They are professional epidemiologists.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
