wessimo said:
Yikes, numbers not really looking great. Will be interesting to see what the curve does over the next couple of weeks.
Projections - like everybody expected - are certainly going up as the country opens up.
Daily Charts
595,980 Views |
2786 Replies |
Last: 2 yr ago by AggieUSMC
Projections - like everybody expected - are certainly going up as the country opens up.
It's mainly the panhandle outbreaks, along with a bit of a spike in Dallas.
The Austin+SA / Houston / DFW portion is still below their mid-April peak. What's changed is the rest of the state.
Notably, the spike last week in FW seems to have subsided a bit.
The Austin+SA / Houston / DFW portion is still below their mid-April peak. What's changed is the rest of the state.
Notably, the spike last week in FW seems to have subsided a bit.
Again, raw numbers are meaningless without context. The more we increase testing the more cases we will identify.PJYoung said:wessimo said:
Yikes, numbers not really looking great. Will be interesting to see what the curve does over the next couple of weeks.
Projections - like everybody expected - are certainly going up as the country opens up.
IHME just updated their projections. Went from 72k by August 4th to 134k by August 4th. Putting it more in line with most other models.HotardAg07 said:
It appears that the deaths data is lifting off the backend of the IHME model.
This model from https://covid19-projections.com/ appears to be doing a better job of predicting the uncertainty on the tail.
Day of the week trend basically in line with the last 4 weeks.
I like watching this one because in NYC they are tracking deaths on the day it happened, not on the day it was reported. As you can see here, the peak in actual deaths in NYC was about 17 days after the lock down. However, the reported peak looks later due to the lag in reporting.
Not good news on the numbers, but glad to see the model the Whitehouse prefers catching up to reality.
The new IHME model projects 968 deaths in Texas by May 4th, with a range of 931-1030.
The actual number is 884.
How is that catching up to reality?
The actual number is 884.
How is that catching up to reality?
Keegan99 said:
The new IHME model projects 968 deaths in Texas by May 4th, with a range of 931-1030.
The actual number is 884.
How is that catching up to reality?
Thats a pretty damn minor deviation in the modeling. They're not prescribing it what it should be, they're running it again from the start with some of the variables accounted for since we know what they are. Or at least that's how I'd approach it. Being off by ~50-150 is pretty meaningless, IMO. Focus on orders of magnitude.
Winner winner chicken dinner.MBAR said:Keegan99 said:
The new IHME model projects 968 deaths in Texas by May 4th, with a range of 931-1030.
The actual number is 884.
How is that catching up to reality?
Thats a pretty damn minor deviation in the modeling. They're not prescribing it what it should be, they're running it again from the start with some of the variables accounted for since we know what they are. Or at least that's how I'd approach it. Being off by ~50-150 is pretty meaningless, IMO. Focus on orders of magnitude.
Sure, if you want to give them a full decimal point, since we're between 100 and 10,000 deaths, I guess it's good?
Realty vs their center-line projection is off by roughly 10% right out of the gate, and reality is solidly outside their window of confidence.
Their model included all data through May 1st. The model vastly overshot a mere 72 hours into the future.
But if you want to see really bad modeling, look at their Wyoming forecast. Wyoming has had 7 deaths for well over a week. The IHME model is predicting 6 total deaths for Wyoming through August 4th, which would mean -1 deaths over the next three months.
That could be considered an improvement from a two weeks ago when the model was predicting 250+ deaths for Wyoming by the end of this month.
Realty vs their center-line projection is off by roughly 10% right out of the gate, and reality is solidly outside their window of confidence.
Their model included all data through May 1st. The model vastly overshot a mere 72 hours into the future.
But if you want to see really bad modeling, look at their Wyoming forecast. Wyoming has had 7 deaths for well over a week. The IHME model is predicting 6 total deaths for Wyoming through August 4th, which would mean -1 deaths over the next three months.
That could be considered an improvement from a two weeks ago when the model was predicting 250+ deaths for Wyoming by the end of this month.
I mean, I know a guy who beat death once.....
perfect example of trying to make models do what they aren't designed to do. texas is a good example, and wyoming a great example of why complex statistics based modeling is incredibly difficult on small data sets. with small numbers and data sets one minor variation causes a large amount of noise in the data. the bigger the data set, the smoother that noise gets and the better modeling and forecasts gets. take this hympotehtical scenario. for 4 days in a row you have the following total cases:Keegan99 said:
Sure, if you want to give them a full decimal point, since we're between 100 and 10,000 deaths, I guess it's good?
Realty vs their center-line projection is off by roughly 10% right out of the gate, and reality is solidly outside their window of confidence.
Their model included all data through May 1st. The model vastly overshot a mere 72 hours into the future.
But if you want to see really bad modeling, look at their Wyoming forecast. Wyoming has had 7 deaths for well over a week. The IHME model is predicting 6 total deaths for Wyoming through August 4th, which would mean -1 deaths over the next three months.
That could be considered an improvement from a two weeks ago when the model was predicting 250+ deaths for Wyoming by the end of this month.
- day 1: 1
- day 2: 2
- day 3: 4
- day 4: 8
you have to look at these numbers through a qualitative lens and decide, did the amount the model was wrong cause the decision on how to proceed change. in both the Wyoming and the Texas examples, regardless of which side of the decision to open things back up you are on, I don't see the modeling error to be significant enough to change that decision. in both cases, they were "in the ballpark". from a policy making level 250 deaths and 7 deaths across an entire state are functionally equivalent. same thing in Texas
Those 15 states are Alaska, Wyoming, Montana, Hawaii, Vermont, Maine, West Virginia, North Dakota, Idaho, New Hampshire, South Dakota, Oregon, Oklahoma, Utah, and Arizona.johnnyblaze36 said:
15 states had ZERO deaths today. It's time to get back to life as usual. Let any business that wants to open do so and all sports leagues should resume play by the weekend.
I'm not saying I wouldn't start opening things back up. I'm saying if the situation in New York shouldn't dictate policy elsewhere, the situation in nearly all of those states shouldn't dictate policy elsewhere.
Arizona is the only semi-big one with any type of sprawling metropolis. Maybe Portland, too.
Today, Arizona reported more deaths than they have on any other day.
Looks like they doubled up USA total projected deaths:
https://covid19.healthdata.org/united-states-of-america
https://covid19.healthdata.org/united-states-of-america
Premium said:
Looks like they doubled up USA total projected deaths:
https://covid19.healthdata.org/united-states-of-america
That 60-70K was never going to happen.
they changed their model methodology. You can expect it will get changed again in the future.
What do you boys want for breakfast BBQ ?.....OK Chili.
Bruce Almighty said:Premium said:
Looks like they doubled up USA total projected deaths:
https://covid19.healthdata.org/united-states-of-america
That 60-70K was never going to happen.
Especially since it already has. I'm thinking more like 200,000 by end of July, but who knows. A lot of that model assumes that people will make good choices. I have teenagers and I read texags, so my view is a bit cynical.
We've hit over 70k, right?Bruce Almighty said:Premium said:
Looks like they doubled up USA total projected deaths:
https://covid19.healthdata.org/united-states-of-america
That 60-70K was never going to happen.
"If you run into an ******* in the morning, you ran into an *******. If you run into *******s all day, you're the *******." – Raylan Givens, "Justified."
Doesn't Wyoming have roughly the same population as Fort Worth?
Confirmed Active Cases Around Texas (5.09.20)
If you live in a green county, you have five or fewer known active cases in your county. Your a phase ahead of the rest of the state.
If you live in a green county, you have five or fewer known active cases in your county. Your a phase ahead of the rest of the state.
- Yellow: 6-100 active cases
- Orange: 100-200 active cases
- Red: 200-400 active cases
- Purple: 400-800 active cases
- Black: 800+ active cases
Texas has started publishing estimated active case data, i.e. confirmed cases less estimated recovered and deaths.
Per the State health department website formula for estimates is as follows:
Per the State health department website formula for estimates is as follows:
- Total Confirmed Cases
- Less: Any Deaths
- Of the remaining, estimate 20% will require hospitalization / 80% will not
- Hospitalized case recovery time is ~32 days
- Non-hospitalized case recovery time is ~14 days
LOL wonder what kind of bs can come out of comparing new "estimated" cases based on an unbiased "formula".
County-level data are pretty interesting...
DALLAS/FT. WORTH AREA
HOUSTON AREA
I-35 CORRIDOR
PANHANDLE
BORDER COUNTIES
DALLAS/FT. WORTH AREA
HOUSTON AREA
I-35 CORRIDOR
PANHANDLE
BORDER COUNTIES
To clarify, the new cases are confirmed tests reported to the state. The active cases are net of the estimated recovered cases, which is dramatically lower than total confirmed cases and provides a relatively better real-time look at where counties are. It's pretty much impossible to track when every confirmed case recovers, so the estimate is as good as it's gonna get.
This last one struck me as strange. Fatality rates as a % of total known infections as been floating around 2.75%. It should be closer to ~1% based on global data. Probably means we're not testing enough, i.e. the denominator is too low.
That said, testing is also really ramping up. To date, Texas has performed and received results for 343,700 tests. There are an additional 133,400 pending tests right now, or and additional ~40% on top of those reported to date.
That said, testing is also really ramping up. To date, Texas has performed and received results for 343,700 tests. There are an additional 133,400 pending tests right now, or and additional ~40% on top of those reported to date.
Fitch said:
Confirmed Active Cases Around Texas (5.09.20)
If you live in a green county, you have five or fewer known active cases in your county. Your a phase ahead of the rest of the state.
- Yellow: 6-100 active cases
- Orange: 100-200 active cases
- Red: 200-400 active cases
- Purple: 400-800 active cases
- Black: 800+ active cases
This is the way it needs to be tracked from now on.... Active Cases
I'm in the Green BTW
Yeah, that's badass. Would like to see an animation of active cases from February to now
Why is Walker County red?
I am pretty sure they have a nursing home outbreak.agz win said:
Why is Walker County red?
No material on this site is intended to be a substitute for professional medical advice, diagnosis or treatment. See full Medical Disclaimer.
agz win said:
Why is Walker County red?
2 above yours
Red: 200-400 active cases
Our percent of tests that are positive has dropped from ~10% to about ~6%. Doubling the number of tests didn't turn up very many additional positives. I don't think increased testing is going to get you where you want to go.Fitch said:
This last one struck me as strange. Fatality rates as a % of total known infections as been floating around 2.75%. It should be closer to ~1% based on global data. Probably means we're not testing enough, i.e. the denominator is too low.
That said, testing is also really ramping up. To date, Texas has performed and received results for 343,700 tests. There are an additional 133,400 pending tests right now, or and additional ~40% on top of those reported to date.
Interesting (but unsurprising) correlation pattern
How long will it take to return results on all those pending tests? We could be in a lock down for months waiting on results from a month ago for people who aren't even sick anymore.
To your point, Texas just about doubled the daily tests and halved the % positive rate. To me that would suggest there's a relatively steady-state of case increases (flat growth or R(t)= ~1).
They've been taking anywhere from 2-14 days to turn around, though reportedly they've been doing better about getting turnaround within a week. It's pretty much a crap shoot from what I can tell.
The problem with a rolling average of hospitalizations is that if length of stay increases a bit, the number gets inflated. It's a good metric for tracking hospital capacity, but not for tracking the number of people needing hospitalization.
FWIW, the last four Monday morning censuses from DSHS:
4/20 - 1471
4/27 - 1542
5/4 - 1540
5/11 - 1525
That's flat for two weeks.
FWIW, the last four Monday morning censuses from DSHS:
4/20 - 1471
4/27 - 1542
5/4 - 1540
5/11 - 1525
That's flat for two weeks.