- Joined
- 10 Dec 06
- Moves
- 8528
@Eladar
Yeah, there is always going to be some mismatch data in an operation of this magnitude. How does someone with such limited capabilities get around that in a consistent way.
I would like to know if anyone has a hypothesis as to what potential features of the IHME models drives the asymmetry in Deaths Per Day across the inflection point? Understanding that seems to be the key as to why our models will underestimate against IHME projections. It must be Hospitalization Driven ( as that is the last remaining part of the model, and it posses the same asymmetry. But why is that the case?
PS I notice you have "3 Alert Moderators !" on your last post? I have noticed rather large numbers of them on many of your other posts, that don't seem violate forum guidelines to me. Is that something you personally can see? Do I have a number of "Alert Moderators !" on my post as well?
- Joined
- 10 Dec 06
- Moves
- 8528
@eladar saidIts obviously not an ideal situation, but "Worldometers" can do whatever they want. As for the States, reporting standards should have been specified by the Federal Government. Obviously a state level change indicates a lack of directive. Changing the scoring system mid game has to have negative consequences in the accuracy of live data analysis. Its certainly bad practice as far as I know? I think if you have entered this far into it under a certain data collection system, keep doing what you are doing and correct the data afterward, but I'm not a scientist.
Is it honest to change how deaths are counted midstream?
Open Question: What do the scientist think?
@ponderable saidThis is just office policy. The Office of National Statistics gives provisional data as early as they can. The information is subject to change. They probably have a similar lag as Germany for final data. I think it just means German officialdom don't release their statistics until they're finalised.
Then the UK is faster than the Federal republic of Germany. Here our most recent numbers are from Nov. 2019 (https://www-genesis.destatis.de/genesis/[WORD TOO LONG] ...[text shortened]... normal fluctuation...But if you have Weekly numbers for the UK it could already Show as significant.
@joe-shmo saidThey should start another time series. They should report confirmed covid-19 deaths as before and have a separate time series for suspected covid-19 deaths. This is generally good practice when one is changing the way data is collected as one needs some overlap of the data to see what the change means. Is a jump in covid-19 deaths an artifact of the reporting procedure or is it reflective of a real effect?
Its obviously not an ideal situation, but "Worldometers" can do whatever they want. As for the States, reporting standards should have been specified by the Federal Government. Obviously a state level change indicates a lack of directive. Changing the scoring system mid game has to have negative consequences in the accuracy of live data analysis. Its certainly bad practic ...[text shortened]... correct the data afterward, but I'm not a scientist.
Open Question: What do the scientist think?
I think that seeing the number of "Alert Moderator" signs is a subscriber feature.
- Joined
- 10 Dec 06
- Moves
- 8528
@deepthought said"They should start another time series. They should report confirmed covid-19 deaths as before and have a separate time series for suspected covid-19 deaths. This is generally good practice when one is changing the way data is collected as one needs some overlap of the data to see what the change means. Is a jump in covid-19 deaths an artifact of the reporting procedure or is it reflective of a real effect?"
They should start another time series. They should report confirmed covid-19 deaths as before and have a separate time series for suspected covid-19 deaths. This is generally good practice when one is changing the way data is collected as one needs some overlap of the data to see what the change means. Is a jump in covid-19 deaths an artifact of the reporting procedure ...[text shortened]... a real effect?
I think that seeing the number of "Alert Moderator" signs is a subscriber feature.
Ok, so I guess the question is that being done.
"I think that seeing the number of "Alert Moderator" signs is a subscriber feature."
Ahh, I didn't realize that. I am seeing several of Eladar's posts with very high numbers. We rarely every see up/down thumbing that gets above a four count. I know you think I'm a conspiracy theorist, but it seems like a group of people collaborating to have him silenced. Whatever Eladar says I believe he has the right to say it. Its very childlike behavior to try and prevent it in that manner, instead of just attacking the substance.
@joe-shmo saidI'm basing my statement on the way I can't see them. It's not obvious to me why anyone thinks that what he's said is anything that the moderators would take an interest in.
"They should start another time series. They should report confirmed covid-19 deaths as before and have a separate time series for suspected covid-19 deaths. This is generally good practice when one is changing the way data is collected as one needs some overlap of the data to see what the change means. Is a jump in covid-19 deaths an artifact of the reporting procedure o ...[text shortened]... ry childlike behavior to try and prevent it in that manner, instead of just attacking the substance.
- Joined
- 10 Dec 06
- Moves
- 8528
@deepthought saidRight, that is why I thought they were rather suspicious. 9 "Alert Moderators" on an innocuous statement seemed rather fishy. I don't know what they hope to accomplish. But oh well...
I'm basing my statement on the way I can't see them. It's not obvious to me why anyone thinks that what he's said is anything that the moderators would take an interest in.
- Joined
- 10 Dec 06
- Moves
- 8528
Including US data from 15 to 48 (March 15 - April 17) - Worldometers
The quadratic regression coefficients for the Gaussian Distribution:
b, c, d
-0.005486517, 0.504025546, -3.779000319
The distribution coefficients:
m, a, A
0.005486517, 45.93310515, 2433
R² = 0.976
The Cumulative Projected Deaths:
Int[g(x)dx] = G( ∞ ) = 58,210
The latest IHME regression 60,308.
It is noticeable now that IHME has given up on the failing hospitalization factor that was driving the asymmetry and thus driving the totals up. Deaths Per day and Hospitalization Resources now possess a symmetry by inspection. Our comparisons now seem to be converging.
https://covid19.healthdata.org/united-states-of-america
<code>∫</code>
@joe-shmo saidI don't know about their model per se. If one attempted to predict infections and used the number of infections to predict the number of deaths then the median time from hospitalisation to admittance to ICU is about a day and the median time to die after admittance to ICU is 6 days. There is a tail, so some people die a couple of months after being infected. This is going to tend to make the deaths curve asymmetric. IMHE takes this into account somehow.
@Eladar
Yeah, there is always going to be some mismatch data in an operation of this magnitude. How does someone with such limited capabilities get around that in a consistent way.
I would like to know if anyone has a hypothesis as to what potential features of the IHME models drives the asymmetry in Deaths Per Day across the inflection point? Understanding that see ...[text shortened]... at something you personally can see? Do I have a number of "Alert Moderators !" on my post as well?