Coronavirus outbreak

[stowie]

I have know idea all I was pointing out is which ever app you use that is not centralised. They all run off the Apple/Google close contact program and are basically the same. So they will all have issues around false positives.

The Google / Apple common API has a couple of ways to access data - from my limited understanding of the Google API documentation. The app developer can either receive a precalculated risk score which uses the Google / Apple code to manage via whatever algorithms they have implemented (not sure what these are, the documentation is very coy about it!). Or the app developer can take the "raw data" and manage the exposure calculation within their app. This allows the developer to implement their own methodology to best capture true positives etc.

This means apps based on the GAEN API may have exactly the same exposure calculations, or they may not depending upon implementation. Even if they use the precalculated risk score they still have various things they could do with this data that would make the apps function differently - for example they could set thresholds differently or weight the exposures in different ways even after the common API has had its go with it.

The fundamental problem is this :

Attenuation is a very noisy proxy of distance. A very low attenuation will indicate a very high probability of a short distance, but a low attenuation can be caused by many phenomena and is not always indicative of a long distance. When setting attenuationDurations' thresholds or risk score multipliers, it's necessary to make a tradeoff between precision and recall.

This aren't my words, but cut and paste from the Google API documentation itself.

What this means is that if two people are standing face-face with their phones in front of them, the attenuation will be very low and the app can calculate with good certainty that they are in close proximity. However, if both people are still standing face-face but their phones are in their back pockets the attenuation may be very different (humans absorb the 2.5Ghz frequency BT operates on rather well). In this case the higher attenuation may mean they are in close proximity with an attenuating environment or they could be far away in an environment which isn't as attenuating. And this is a really, really simple case. Common environments are much more complex as studies have shown.

So the app developers and Google/Apple have a difficult task to get measurements which minimise false results.

This doesn't mean the app is useless - far from it. But it is one of a number of tools that have to be employed to effectively isolate COVID cases and minimise spread. It isn't a panacea.

 

[marinyork]

The Google / Apple common API has a couple of ways to access data - from my limited understanding of the Google API documentation. The app developer can either receive a precalculated risk score which uses the Google / Apple code to manage via whatever algorithms they have implemented (not sure what these are, the documentation is very coy about it!). Or the app developer can take the "raw data" and manage the exposure calculation within their app. This allows the developer to implement their own methodology to best capture true positives etc.

This means apps based on the GAEN API may have exactly the same exposure calculations, or they may not depending upon implementation. Even if they use the precalculated risk score they still have various things they could do with this data that would make the apps function differently - for example they could set thresholds differently or weight the exposures in different ways even after the common API has had its go with it.

I see you are still knocking around.

The issue here is entirely the family of functions of calculations they could employ. You do have form for continuously writing trivial things and then arguing with others who you obtusely wish to misunderstand.

The last paragraph you have written is correct.

There are papers that go through the 'not sure' algorithms if you were interested.

 

[stowie]

I see you are still knocking around.

The issue here is entirely the family of functions of calculations they could employ. You do have form for continuously writing trivial things and then arguing with others who you obtusely wish to misunderstand.

The last paragraph you have written is correct.

There are papers that go through the 'not sure' algorithms if you were interested.

I have repeatedly asked you for details on the papers because I am interested. Still waiting.

 

[tom73]

The Google / Apple common API has a couple of ways to access data - from my limited understanding of the Google API documentation. The app developer can either receive a precalculated risk score which uses the Google / Apple code to manage via whatever algorithms they have implemented (not sure what these are, the documentation is very coy about it!). Or the app developer can take the "raw data" and manage the exposure calculation within their app. This allows the developer to implement their own methodology to best capture true positives etc.

This means apps based on the GAEN API may have exactly the same exposure calculations, or they may not depending upon implementation. Even if they use the precalculated risk score they still have various things they could do with this data that would make the apps function differently - for example they could set thresholds differently or weight the exposures in different ways even after the common API has had its go with it.

The fundamental problem is this :



This aren't my words, but cut and paste from the Google API documentation itself.

What this means is that if two people are standing face-face with their phones in front of them, the attenuation will be very low and the app can calculate with good certainty that they are in close proximity. However, if both people are still standing face-face but their phones are in their back pockets the attenuation may be very different (humans absorb the 2.5Ghz frequency BT operates on rather well). In this case the higher attenuation may mean they are in close proximity with an attenuating environment or they could be far away in an environment which isn't as attenuating. And this is a really, really simple case. Common environments are much more complex as studies have shown.

So the app developers and Google/Apple have a difficult task to get measurements which minimise false results.

This doesn't mean the app is useless - far from it. But it is one of a number of tools that have to be employed to effectively isolate COVID cases and minimise spread. It isn't a panacea.

I never thought it was though it has some merit it's not without it's issues. As for being part of any effective response it's not and is mostly window dressing. To take away from the real issue that testing is a total mess and the only real way to do contact tracing is on a local level.
By people who know what they are doing and know what the word is street. They know what to they've done it for years and it works.
Which is more than can be said for Dido's costly horror show. The App It's yet another example of "machine that go's ping" health care.

 

[fossyant]

Both main Manchester Universities are now online teaching from tomorrow, unless some lessons have to be practicals.

 

[DCLane]

Both main Manchester Universities are now online teaching from tomorrow, unless some lessons have to be practicals.

That's exactly what I'm doing until Christmas. They're coming in for 2 hours a fortnight only. My module's completely online although we're delivering more than normal having added 50% more to the delivery time.

 

[Ming the Merciless]

It just gets better and better, their phone lines crash.

https://apple.news/A08AD5mtnTbuu_zD81ENdEw

 

[roubaixtuesday]

Just in case you thought it was a blip, yet another record, doubling time looks consistent with ~10 days, rapidly catching up with European neighbours.

Hospitalizations, deaths on same exponential growth trajectory.

Is this reminding people of anything?

 

[tom73]

@roubaixtuesday we really have learned nothing.
In the mean time in New Zealand Auckland having got back in control of local outbreak. Yesterday after recording day 9 case free are now again fully open and life is normal.

 

[Oldhippy]

Profit before life sadly, and ignorant people who really don't seem to think it will be them but expect all the care going if they end up in hospital.

 

[marinyork]

Contact tracer I know made 40+ calls today. Previous months it was 1 call a month and in recent days 5-20 calls a day.

 

[tom73]

Contact tracer I know made 40+ calls today. Previous months it was 1 call a month and in recent days 5-20 calls a day.

All 1st contact or following up on no pick up ? Or more of a mix ?

 

[marinyork]

All 1st contact or following up on no pick up ? Or more of a mix ?

Tier 3. She/he makes the calls telling people to self isolate, as does everyone on her/his team.

Interestingly enough the team made a number of calls last week telling people to self isolate after 14 days had expired.

 

[marinyork]

All 1st contact or following up on no pick up ? Or more of a mix ?

All first contact. The no pick ups didn't say, although last week it sounded like it was about 80% no pick ups.

Of those they got through to, students and many people whose first language is not English.

 

[tom73]

Tier 3. She/he makes the calls telling people to self isolate, as does everyone on her/his team.

Interestingly enough the team made a number of calls last week telling people to self isolate after 14 days had expired.

Right with you , I've heard about contacting after it had pasted before. It was also talked about in the last panorama about T&T.