The predictive value of COVID positivity

[Not_a_Number]

1 minute ago, lovelearnandlive said:

Thurs, Nov 26 +1443, 7-day 1623

Fri, Nov 27 +1356, 7-day 1532

Sat, Nov 28 +1224, 7-day 1491

Sun, Nov 29 +820, 7-day 1483

Mon, Nov 30 +1275, 7-day 1522

Tues, Dec 1 +2667, 7-day 1589

Wed, Dec 2 +2874, 7-day 1665

Thurs, Dec 3 +2926, 7-day 1878

Fri, Dec 4 +2738, 7-day 2074

Sat, Dec 5 +2266, 7-day 2224

Sun, Dec 6 +1089, 7-day 2262

Mon, Dec 7 +1508, 7-day 2296

Tues, Dec 8 +2913, 7-day 2337

Wed, Dec 9 +3243, 7-day 2398

 

It looks like we probably crossed 3000 even accounting for the fact that CO’s numbers are high. Not a great day either way. Besides CO, 4 other states posted over 200 deaths (almost 300 for TX). 😞

@Danae has been running the numbers, and case numbers may, indeed, be the best correlate of deaths at this point 😞 . In which case we're going to have an even worse December than I thought... 😕 

 

 

[Not_a_Number]

The two-week average ignoring the data dumps is now around 2000, which is... really not good. 

It looks like right now the right number is somewhere between cases and positivity, which is rather hard to work with. So, @Corraleno, that means I was probably too optimistic and you were probably too pessimistic. 

[The Governess]

5 minutes ago, Not_a_Number said:

The two-week average ignoring the data dumps is now around 2000, which is... really not good. 

It looks like right now the right number is somewhere between cases and positivity, which is rather hard to work with. So, @Corraleno, that means I was probably too optimistic and you were probably too pessimistic. 

Where do the two models say we will be two weeks from now?

[Not_a_Number]

33 minutes ago, lovelearnandlive said:

Where do the two models say we will be two weeks from now?

Let's see. Cases say 3600 deaths a day on average (eek), positivity says 2500 deaths a day on average. 

Just to be clear, both models say that things will really suck, lol. But there's a difference. 

[The Governess]

4 minutes ago, Not_a_Number said:

Let's see. Cases say 3600 deaths a day on average (eek), positivity says 2500 deaths a day on average. 

Just to be clear, both models say that things will really suck, lol. But there's a difference. 

Both really do suck. But wow, that 3600 number puts us at 25,000 dying per week by Christmas. 😢

[GoodGrief3]

2 hours ago, Not_a_Number said:

Lol. Yeah, I wish they'd let us know, too!! Probably not recent data, either way, though. Data dumps never are... 

My husband (who works in the field) told me last week that 1000 missed cases just showed up for our state (which has relatively low numbers). I haven't seen them show up in the daily case counts yet and am curious about how it will be handled. Maybe just one dramatically high day.

[Corraleno]

If I use the case average that was actually reported on Friday 11/27, 170K x 0.017 = 2920 next week, but it makes more sense to use the average for the Wednesday before Thanksgiving, due to the artificial dip on the 26th & 27th. So 180K x 0.017 = 3060 (average) deaths by the end of next week.

Case average on 12/4 was 187K, which gives a prediction of around 3180 by Christmas. 

Using the current case average (213K) to predict 3 weeks from now, a CFR of 1.7 = 3600 (or 1.6 CFR = 3400) by New Years Eve. However, if the current case average includes some large data dumps, that could artificially inflate the predicted death average.  Plus there will be a big data lag over Christmas, and then probably an even slower and more drawn out catch-up period because of continuing lags between Christmas and New Year, so the numbers are likely to be all over the place before and after Christmas. Which could make the first week of January look seriously scary as the backlog gets cleared. 

[Corraleno]

6 hours ago, Not_a_Number said:

Let's see. Cases say 3600 deaths a day on average (eek), positivity says 2500 deaths a day on average. 

Just to be clear, both models say that things will really suck, lol. But there's a difference. 

 

6 hours ago, lovelearnandlive said:

Both really do suck. But wow, that 3600 number puts us at 25,000 dying per week by Christmas. 😢

Actually the prediction of 3600 is for three weeks from now (New Year's), not two weeks (Christmas). Case average on 12/4 was 187K x 0.017 = 3180 three weeks later on 12/25.

[Not_a_Number]

2 hours ago, Corraleno said:

 

Actually the prediction of 3600 is for three weeks from now (New Year's), not two weeks (Christmas). Case average on 12/4 was 187K x 0.017 = 3180 three weeks later on 12/25.

Yeah, sorry. I was using 2 weeks and keep forgetting you use 3.

[Not_a_Number]

2 hours ago, Corraleno said:

However, if the current case average includes some large data dumps, that could artificially inflate the predicted death average.

I don’t know if it does. I haven’t seen huge data dumps for cases, just deaths. Even those haven’t been giant.

[Ausmumof3]

10 hours ago, Not_a_Number said:

The two-week average ignoring the data dumps is now around 2000, which is... really not good. 

It looks like right now the right number is somewhere between cases and positivity, which is rather hard to work with. So, @Corraleno, that means I was probably too optimistic and you were probably too pessimistic. 

I think that model makes the most sense though logically?  I mean positivity can be impacted by the approach to testing and the number of cases can be detected by the approach to testing but their should be a relationship between the two and the actual number of cases (versus the number of cases we’re detecting).  Though I have no idea how to turn that into a formula...

[Not_a_Number]

19 minutes ago, Ausmumof3 said:

I think that model makes the most sense though logically?  I mean positivity can be impacted by the approach to testing and the number of cases can be detected by the approach to testing but their should be a relationship between the two and the actual number of cases (versus the number of cases we’re detecting).  Though I have no idea how to turn that into a formula...

I think when testing is scarce enough, positivity is by far the best statistic. But we’re obviously not there anymore... at least for now.

[Not_a_Number]

Things were relatively stable from last week today. Here's my attempt to draw the "actual curve": 

[The Governess]

Thurs, Nov 26 +1443, 7-day 1623

Fri, Nov 27 +1356, 7-day 1532

Sat, Nov 28 +1224, 7-day 1491

Sun, Nov 29 +820, 7-day 1483

Mon, Nov 30 +1275, 7-day 1522

Tues, Dec 1 +2667, 7-day 1589

Wed, Dec 2 +2874, 7-day 1665

Thurs, Dec 3 +2926, 7-day 1878

Fri, Dec 4 +2738, 7-day 2074

Sat, Dec 5 +2266, 7-day 2224

Sun, Dec 6 +1089, 7-day 2262

Mon, Dec 7 +1508, 7-day 2296

Tues, Dec 8 +2913, 7-day 2337

Wed, Dec 9 +3243, 7-day 2392

Thurs, Dec 10 2974, 7-day 2407

 

Today’s numbers... they don’t match up exactly - it looks like worldometers revised some numbers (upward slightly). I’m too tired to go back and edit, but it looks like whatever they did bumped to averages up by around 10 (yesterday’s average is now showing as 2401 rather than 2382).

 

We will break 300,000 deaths tomorrow. 😞

 

[Roadrunner]

“Los Angeles County Director of Health and Human Services Dr. Christina Ghaly has laid out the following equation, which has proven reliable: Ghaly has that about 12% of all coronavirus cases end up in the hospital. “Half of those end up in ICU,” she said in November. “Two-thirds of those are on a ventilator. Half of those will die, based on previous experience.”

https://www.yahoo.com/entertainment/california-covid-19-wednesday-marks-211819432.html

 

Edited December 11, 2020 by Roadrunner

[The Governess]

10 hours ago, Roadrunner said:

“Los Angeles County Director of Health and Human Services Dr. Christina Ghaly has laid out the following equation, which has proven reliable: Ghaly has that about 12% of all coronavirus cases end up in the hospital. “Half of those end up in ICU,” she said in November. “Two-thirds of those are on a ventilator. Half of those will die, based on previous experience.”

https://www.yahoo.com/entertainment/california-covid-19-wednesday-marks-211819432.html

 

And our cases haven’t peaked yet. I’m very concerned. 

[cintinative]

14 hours ago, Roadrunner said:

“Los Angeles County Director of Health and Human Services Dr. Christina Ghaly has laid out the following equation, which has proven reliable: Ghaly has that about 12% of all coronavirus cases end up in the hospital. “Half of those end up in ICU,” she said in November. “Two-thirds of those are on a ventilator. Half of those will die, based on previous experience.”

https://www.yahoo.com/entertainment/california-covid-19-wednesday-marks-211819432.html

 

My husband has been asking about statistics regarding how many become really ill, need hospitalization, etc. Do you think these figures are a good reference? Or is there a better one somewhere? Thank you.

 

[Roadrunner]

21 minutes ago, cintinative said:

My husband has been asking about statistics regarding how many become really ill, need hospitalization, etc. Do you think these figures are a good reference? Or is there a better one somewhere? Thank you.

 

I don’t know. These are from La county. I would think it’s representative. Sadly. 

[Not_a_Number]

31 minutes ago, cintinative said:

My husband has been asking about statistics regarding how many become really ill, need hospitalization, etc. Do you think these figures are a good reference? Or is there a better one somewhere? Thank you.

 

I don’t think that’s overall right. I’ve seen much lower stats for hospitalization before.

[Lecka]

It seems like the positivity rate for my state jumps around a lot, and will vary by a huge amount depending on the day and who is calculating it — does this happen for other states?  I’m in Oklahoma.

Like — I will go from seeing 8% to seeing 21% and it seems like it’s in a matter of 2-3 days.  
 

 

[Pam in CT]

re newly released DHS dataset

1 hour ago, cintinative said:

My husband has been asking about statistics regarding how many become really ill, need hospitalization, etc. Do you think these figures are a good reference? Or is there a better one somewhere? Thank you.

 

Just as of this week, DHS has been publishing hospital-level data and that same data aggregating to the county level, back through early August, and various folks have been starting to work out ways of presenting it, including this interactive map that NPR built that lets you see by county what percentage of total hospital beds are currently occupied by COVID patients.

The University of Minnesota has the full data set up with many more metrics, and has been working on other forms of visualizations.

[The Governess]

Regarding hospitalizations as a percent of cases...

Back in late Sept/early Oct in LA county, about 1,000 cases per day on average were being reported. During this time hospitalizations were pretty consistent, around 750 total including ICU. 

At the end of November, there was a big jump to about 4,000 cases per day on average. Hospitalizations are around 3500 now and still increasing. They are increasing about 700 per week but that number would include new admissions less discharges and deaths. 

7-day average new cases for the county now are 9,000. If this trend holds could hospitalizations increase to 7-8,000 in a couple weeks or am I missing something? Please tell me I’m missing something.

[Not_a_Number]

16 minutes ago, lovelearnandlive said:

Regarding hospitalizations as a percent of cases...

Back in late Sept/early Oct in LA county, about 1,000 cases per day on average were being reported. During this time hospitalizations were pretty consistent, around 750 total including ICU. 

At the end of November, there was a big jump to about 4,000 cases per day on average. Hospitalizations are around 3500 now and still increasing. They are increasing about 700 per week but that number would include new admissions less discharges and deaths. 

7-day average new cases for the county now are 9,000. If this trend holds could hospitalizations increase to 7-8,000 in a couple weeks or am I missing something? Please tell me I’m missing something.

It depends if they are catching the same kinds of cases or not. It’s hard to know — they could be catching milder cases on average, or more severe ones.

[Not_a_Number]

33 minutes ago, Lecka said:

It seems like the positivity rate for my state jumps around a lot, and will vary by a huge amount depending on the day and who is calculating it — does this happen for other states?  I’m in Oklahoma.

Like — I will go from seeing 8% to seeing 21% and it seems like it’s in a matter of 2-3 days.  
 

 

Positivity is spiky... best to use the 7-day averages:

https://coronavirus.jhu.edu/testing/individual-states/oklahoma

 

[cintinative]

3 hours ago, Not_a_Number said:

Positivity is spiky... best to use the 7-day averages:

https://coronavirus.jhu.edu/testing/individual-states/oklahoma

 

Wow, their percent positive is way higher than the one our state has put out. Our state says it is around 15.7 % and they say 19.5%

[Pam in CT]

11 minutes ago, cintinative said:

Wow, their percent positive is way higher than the one our state has put out. Our state says it is around 15.7 % and they say 19.5%

JHU includes antigen tests (not antibody, antigen).  I believe that some states do as well, others only count PCRs.

PCR v antigen v antibody tests

[cintinative]

16 minutes ago, Pam in CT said:

JHU includes antigen tests (not antibody, antigen).  I believe that some states do as well, others only count PCRs.

PCR v antigen v antibody tests

You are right. I didn't notice it before but Ohio's does not include the antigens. Thank you.

[The Governess]

Fri, Nov 27 +1356, 7-day 1531

Sat, Nov 28 +1224, 7-day 1491

Sun, Nov 29 +820, 7-day 1482

Mon, Nov 30 +1271, 7-day 1522

Tues, Dec 1 +2666, 7-day 1587

Wed, Dec 2 +2872, 7-day 1665

Thurs, Dec 3 +2924, 7-day 1876

Fri, Dec 4 +2703, 7-day 2069

Sat, Dec 5 +2266, 7-day 2217

Sun, Dec 6 +1100, 7-day 2257

Mon, Dec 7 +1562, 7-day 2299

Tues, Dec 8 +2980, 7-day 2344

Wed, Dec 9 +3261, 7-day 2400

Thurs, Dec 10 3098, 7-day 2424

Friday, Dec 11 3019, 7-day 2469

I added today’s numbers and went back and edited past numbers so they reflect what is in worldometers. There were a few days this week that were revised upward a bit.

[The Governess]

Just another depressing update...

Sun, Nov 29 +820, 7-day 1482

Mon, Nov 30 +1271, 7-day 1522

Tues, Dec 1 +2666, 7-day 1587

Wed, Dec 2 +2872, 7-day 1665

Thurs, Dec 3 +2924, 7-day 1876

Fri, Dec 4 +2703, 7-day 2069

Sat, Dec 5 +2266, 7-day 2217

Sun, Dec 6 +1100, 7-day 2257

Mon, Dec 7 +1562, 7-day 2299

Tues, Dec 8 +2980, 7-day 2344

Wed, Dec 9 +3261, 7-day 2400

Thurs, Dec 10 3098, 7-day 2424

Friday, Dec 11 3019, 7-day 2469

Sat, Dec 12 +2307, 7-day 2478

Sun, Dec 13 1379, 7-day 2519

 

[The Governess]

Today’s update:

Mon, Nov 30 +1271, 7-day 1522

Tues, Dec 1 +2666, 7-day 1587

Wed, Dec 2 +2872, 7-day 1665

Thurs, Dec 3 +2924, 7-day 1876

Fri, Dec 4 +2703, 7-day 2069

Sat, Dec 5 +2266, 7-day 2217

Sun, Dec 6 +1100, 7-day 2257

Mon, Dec 7 +1562, 7-day 2299

Tues, Dec 8 +2980, 7-day 2344

Wed, Dec 9 +3261, 7-day 2400

Thurs, Dec 10 3098, 7-day 2424

Friday, Dec 11 3019, 7-day 2469

Sat, Dec 12 +2307, 7-day 2478

Sun, Dec 13 1379, 7-day 2519

Mon, Dec 14 +1619, 7-day 2527

Only a small increase in the average today.

[Ausmumof3]

3486 deaths today.  I’m guessing some of those are from the weekend.  It does look like the case curve and deaths rolling average are flattening off now.  

[Ausmumof3]

https://www.sun-sentinel.com/coronavirus/fl-ne-ss-prem-covid-deaths-florida-election-20201216-f4kgezjf4rf75ppumt4omxfsxy-story.html?outputType=amp&__twitter_impression=true
 

not totally related but don’t want to start a new thread.  What did people make or this story around Florida’s data?

[The Governess]

Here is an update. We had a day this week where the average seemed to hold steady and then it started rising again. 😞

Thurs, Dec 3 +2924, 7-day 1876

Fri, Dec 4 +2704, 7-day 2069

Sat, Dec 5 +2267, 7-day 2218

Sun, Dec 6 +1109, 7-day 2259

Mon, Dec 7 +1559, 7-day 2299

Tues, Dec 8 +2977, 7-day 2344

Wed, Dec 9 +3263, 7-day 2400

Thurs, Dec 10 3094, 7-day 2424

Friday, Dec 11 3029, 7-day 2469

Sat, Dec 12 +2318, 7-day 2478

Sun, Dec 13 1388, 7-day 2519

Mon, Dec 14 +1622, 7-day 2527

Tues, Dec 15 +2976, 7-day 2528

Wed, Dec 16 +3561, 7-day 2573

Thurs, Dec 17 +3,277, 7-day 2599

 

CA is not looking so great. I’m pretty concerned with our health care availability. Newsom actually mentioned looking overseas for additional staffing. They have relaxed the nurse to patient ratios which helps with capacity but not level of care. Our state is reporting more cases now than all other *countries* in the world except the US and Brazil.

[The Governess]

On 12/16/2020 at 6:51 PM, Ausmumof3 said:

https://www.sun-sentinel.com/coronavirus/fl-ne-ss-prem-covid-deaths-florida-election-20201216-f4kgezjf4rf75ppumt4omxfsxy-story.html?outputType=amp&__twitter_impression=true
 

not totally related but don’t want to start a new thread.  What did people make or this story around Florida’s data?

Yeah, that’s pretty suspicious. 😒

[Roadrunner]

1 hour ago, lovelearnandlive said:

Here is an update. We had a day this week where the average seemed to hold steady and then it started rising again. 😞

Thurs, Dec 3 +2924, 7-day 1876

Fri, Dec 4 +2704, 7-day 2069

Sat, Dec 5 +2267, 7-day 2218

Sun, Dec 6 +1109, 7-day 2259

Mon, Dec 7 +1559, 7-day 2299

Tues, Dec 8 +2977, 7-day 2344

Wed, Dec 9 +3263, 7-day 2400

Thurs, Dec 10 3094, 7-day 2424

Friday, Dec 11 3029, 7-day 2469

Sat, Dec 12 +2318, 7-day 2478

Sun, Dec 13 1388, 7-day 2519

Mon, Dec 14 +1622, 7-day 2527

Tues, Dec 15 +2976, 7-day 2528

Wed, Dec 16 +3561, 7-day 2573

Thurs, Dec 17 +3,277, 7-day 2599

 

CA is not looking so great. I’m pretty concerned with our health care availability. Newsom actually mentioned looking overseas for additional staffing. They have relaxed the nurse to patient ratios which helps with capacity but not level of care. Our state is reporting more cases now than all other *countries* in the world except the US and Brazil.

There seems to be no end to how high CA Covid numbers are getting. Every day I look for a glimmer of hope, and every day it gets worse instead. Everywhere else numbers rise and fall. in CA, they only rise. When are we going to teach the peak???!

[The Governess]

8 minutes ago, Roadrunner said:

There seems to be no end to how high CA Covid numbers are getting. Every day I look for a glimmer of hope, and every day it gets worse instead. Everywhere else numbers rise and fall. in CA, they only rise. When are we going to teach the peak???!

And when we finally do reach the peak then we will still have weeks to go before hospitalizations and deaths hit their peak. 😞

 

[mathnerd]

15 minutes ago, Roadrunner said:

There seems to be no end to how high CA Covid numbers are getting. Every day I look for a glimmer of hope, and every day it gets worse instead. Everywhere else numbers rise and fall. in CA, they only rise. When are we going to teach the peak???!

This seems to be a huge tsunami of infections unleashed by travel and get-togethers related to Thanksgiving. I think that we will continue to see the same trend due to Christmas and New Year's events despite the SIP. I hope the state will get a break in late January, but, then, it would be almost time for the Spring-breakers to party, I suppose 😞

Edited December 18, 2020 by mathnerd

[Roadrunner]

16 minutes ago, mathnerd said:

This seems to be a huge tsunami of infections unleashed by travel and get-togethers related to Thanksgiving. I think that we will continue to see the same trend due to Christmas and New Year's events despite the SIP. I hope the state will get a break in late January, but, then, it would be almost time for the Spring-breakers to party, I suppose 😞

The frustrating part is Thanksgiving travel was all across the country. We have the strictest SIP orders. Yet other parts of the country are nowhere near the wave we are having. I really don’t get it.   

[mathnerd]

21 minutes ago, Roadrunner said:

The frustrating part is Thanksgiving travel was all across the country. We have the strictest SIP orders. Yet other parts of the country are nowhere near the wave we are having. I really don’t get it.   

Allow me to speculate a bit: LA and SoCal did not get affected as bad as NYC earlier on due to the SIP orders. So, there was lesser fear of the virus, perhaps, and the economy in those areas is not doing well. A combination of these could have made more people disregard precautions, maybe? I too don't get how this one area is getting hit so badly (infections and deaths in other parts of CA are also on an alarming upswing).

[Roadrunner]

1 minute ago, mathnerd said:

Allow me to speculate a bit: LA and SoCal did not get affected as bad as NYC earlier on due to the SIP orders. So, there was lesser fear of the virus, perhaps, and the economy in those areas is not doing well. A combination of these could have made more people disregard precautions, maybe? I too don't get how this one area is getting hit so badly (infections and deaths in other parts of CA are also on an alarming upswing).

We are in the northern part in a county that has been purple and shut down the entire time. And yes, it’s still getting worse. There are no tools left in a toolbox now. It feels very desperate. 

[Not_a_Number]

5 hours ago, Roadrunner said:

We are in the northern part in a county that has been purple and shut down the entire time. And yes, it’s still getting worse. There are no tools left in a toolbox now. It feels very desperate. 

I assume people are just tired. Shelter in place for a year is just too much 😞 . 

[Lecka]

I think there is an element of luck, too.  If there is a superspreader one place but not another, it can be possible for that event to be a superspreader event, but for the equivalent superspreader event that luckily didn’t become a superspreader event — well, it just worked out for them.

I think it seems like it is not fair this way, and some places can be luckier or unluckier.

Of course I think precautions make a huge difference.  But then I think there’s also that element of luck or chance.  
 

Well — I’m not totally sure on all the details, but I think it’s getting harder to get tested locally.  I think we have numbers going up, more people exposed or with symptoms and needing a test, and then not enough capacity.  
 

It is SO frustrating.  Because — it means all our numbers are probably off.

And then too — it is also just — delaying everything as far as contact  tracing  (edit which we barely have, but if you first hear about an exposure a week or ten days later because the person had to wait to be tested — gets to be a joke).

I hope I am being paranoid, but probably I’m not.

 

Edited December 18, 2020 by Lecka

[Lecka]

I’m not sure about this, but I think it could effect our positivity rate’s accuracy if it is harder to get a walk-in test, if lines are longer, if you have to go through a doctors office, etc.

I think that will skew away from people who can’t take off work to get a test if they can’t just get in before or after work.

And skew away from people who don’t have a doctor.

I think the people who jump through hurdles to get a test or are available any time, might be less likely to be positive than people who are not as able to jump through hurdles, when it gets into there being lines, limited time slots, etc.  

[Pam in CT]

2 hours ago, Lecka said:

I’m not sure about this, but I think it could effect our positivity rate’s accuracy if it is harder to get a walk-in test, if lines are longer, if you have to go through a doctors office, etc.

I think that will skew away from people who can’t take off work to get a test if they can’t just get in before or after work.

And skew away from people who don’t have a doctor.

I think the people who jump through hurdles to get a test or are available any time, might be less likely to be positive than people who are not as able to jump through hurdles, when it gets into there being lines, limited time slots, etc.  

I do think that's an issue that affects how stable is the "denominator" of the positivity rate.  Self-referred / free testing is still *available* here an in NYC, but lines are markedly longer; so people doing it just cuz, or because they'd like to visit an elderly relative, are less likely to do so.  As those folks self-select out, the rate would presumably rise a bit.

A *lot* of employers around here with IRL work are now requiring weekly tests, so something of a bifurcated testing market has arisen, where you can get tested for free (and wait and wait and wait...) at community health centers, v pay at CVS/ other sites and go through much faster.  Like with congestion pricing on toll roads, folks have mixed feelings about it.

[Not_a_Number]

3 minutes ago, Pam in CT said:

I do think that's an issue that affects how stable is the "denominator" of the positivity rate.  Self-referred / free testing is still *available* here an in NYC, but lines are markedly longer; so people doing it just cuz, or because they'd like to visit an elderly relative, are less likely to do so.  As those folks self-select out, the rate would presumably rise a bit.

A *lot* of employers around here with IRL work are now requiring weekly tests, so something of a bifurcated testing market has arisen, where you can get tested for free (and wait and wait and wait...) at community health centers, v pay at CVS/ other sites and go through much faster.  Like with congestion pricing on toll roads, folks have mixed feelings about it.

Yeah, the sampling keeps changing, making the positivity mean different things. It’s definitely not as reliable as it used to be.

That being said, it’s still a useful statistic, I think. The predictions based on cases haven’t worked that well, either. 

[Danae]

For discussion's sake, here's MN positivity (red), cases (yellow), and deaths (blue).  Positivity and cases shifted 14 days forward.

Positivity was definitely the stronger correlation in the beginning, then for most of the summer positivity and case numbers both worked (since test numbers were fairly stable at that point they were measuring basically the same thing).  But when things started to skyrocket in the fall positivity plateaued but cases and deaths kept going up.  I'm guessing that the change was that the population that was voluntarily testing and the population that was getting sick diverged, so a higher percentage of cases were only tested because the person was sick enough to need to see a doctor.  

Cases and positivity are both headed down, but deaths haven't turned the corner yet. 😞

[Not_a_Number]

3 minutes ago, Danae said:

For discussion's sake, here's MN positivity (red), cases (yellow), and deaths (blue).  Positivity and cases shifted 14 days forward.

Positivity was definitely the stronger correlation in the beginning, then for most of the summer positivity and case numbers both worked (since test numbers were fairly stable at that point they were measuring basically the same thing).  But when things started to skyrocket in the fall positivity plateaued but cases and deaths kept going up.  I'm guessing that the change was that the population that was voluntarily testing and the population that was getting sick diverged, so a higher percentage of cases were only tested because the person was sick enough to need to see a doctor.  

Cases and positivity are both headed down, but deaths haven't turned the corner yet. 😞

That's a useful graph. You can see how positivity used to work best, and how right now, in MN at least, cases work best. 

Is there a way to split up the positivity graph (probably somewhere around the summer) so that the scaling factor for the positivity increases? Because we definitely DID change our sampling radically as colleges started to open. Suddenly, our sampling wasn't drawn largely from people with symptoms anymore. You can see this in your graph, in fact. 

I wonder if your idea of changing the scaling factor for positivity using the scaling for the previous month would yield a much more predictive graph using positivity alone. 

[mathnerd]

16 hours ago, Roadrunner said:

The frustrating part is Thanksgiving travel was all across the country. We have the strictest SIP orders. Yet other parts of the country are nowhere near the wave we are having. I really don’t get it.   

So, officially, they are blaming the residents who ignored health safety during Thanksgiving. 1 in 80! That means at least one in every street!

https://www.newsweek.com/1-80-l-county-residents-has-covid-health-officials-blame-thanksgiving-1555403

In Los Angeles County, one in every 80 county residents is believed to be infected with COVID-19 as transmission rates continue rising in the aftermath of Thanksgiving. According to county health officials, the recent surge in new cases is a result of residents ignoring recommendations to stay home over the Thanksgiving holiday.

[Roadrunner]

12 minutes ago, mathnerd said:

So, officially, they are blaming the residents who ignored health safety during Thanksgiving. 1 in 80! That means at least one in every street!

https://www.newsweek.com/1-80-l-county-residents-has-covid-health-officials-blame-thanksgiving-1555403

In Los Angeles County, one in every 80 county residents is believed to be infected with COVID-19 as transmission rates continue rising in the aftermath of Thanksgiving. According to county health officials, the recent surge in new cases is a result of residents ignoring recommendations to stay home over the Thanksgiving holiday.

Out county recorded 10x increase today of new cases. Some backlog, but not enough 😞 

 

 

Edited December 18, 2020 by Roadrunner

[Not_a_Number]

Looks like deaths actually went down a bit! 

I really need to mess with the "scaling factor" for positivity a bit. From the looks of things, cases are about to be totally useless again -- we still haven't hit last week's predicted death number, and this week's was 3060, which is not even close. And we're supposed to climb up to something like 3770, which I'm just not seeing in the shape of the graph. 

It occurred to me that the issue may be is that we're sampling from sick people... but there are a LOT more sick people in December than in July! So we may need to scale by that. 

But scaling by the observed factor would take of that, as well as the sampling issues. I'm very curious how well that works.