Coronavirus - Data Analysis Thread

Coronavirus - Data Analysis Thread

Author
Discussion

Isaac Hunt

12,911 posts

179 months

Monday 21st December 2020
quotequote all
Elysium said:
4. A great site to compare latest info by country:

https://datagraver.com/corona/#/
Good find.

coanda

2,296 posts

158 months

Monday 21st December 2020
quotequote all
It's essential that all graphs posted can be recreated from source data by anyone with a basic level of spreadsheet knowledge, in order to build confidence in the data sets. Poorly presented graphs or graphs with no source data are not very useful and can discredit something that is otherwise genuine. We see this at every government presentation and throughout all industry.

Data manipulation is underhand and uncalled for, and in some case just dangerous. Let the data speak for itself.

smartypants

50,075 posts

137 months

Monday 21st December 2020
quotequote all
coanda said:
It's essential that all graphs posted can be recreated from source data by anyone with a basic level of spreadsheet knowledge, in order to build confidence in the data sets. Poorly presented graphs or graphs with no source data are not very useful and can discredit something that is otherwise genuine. We see this at every government presentation and throughout all industry.

Data manipulation is underhand and uncalled for, and in some case just dangerous. Let the data speak for itself.
Be lovely if HMG would abide by the same notion smile

Elysium

Original Poster:

10,725 posts

155 months

Monday 21st December 2020
quotequote all
I have no idea how to share the data set behind these graphs, but it is all taken directly from here:

https://coronavirus.data.gov.uk/

My cases per 100k graph compares positive cases by specimen date to tests processed two days later.

1. Tests and Cases. Although cases have been increasing, we have also been testing more, topping 440k tests processed in a day for the first time. This suggests to me that we have reversed the lockdown 'dip', so what happens next will be interesting:



2. Cases per 100k tests, admissions and deaths. Cases and admissions now showing an uptick. Deaths will follow:







3. Overlay of the above and corresponding growth rates:




QuartzDad

1,377 posts

90 months

Monday 21st December 2020
quotequote all
The Government data dashboard now includes 7 day PCR positivity rates which presumably ameliorates the 'more testing, more cases' effect. Only works at nation level currently - this is for England:

https://coronavirus.data.gov.uk/details/testing?ar...



Elysium

Original Poster:

10,725 posts

155 months

Monday 21st December 2020
quotequote all
Well spotted. This is essentially the same curve as my 'cases per 100k tests' graph.

Unfortunately its only for England at the moment.

QuartzDad

1,377 posts

90 months

Monday 21st December 2020
quotequote all
Elysium said:
Well spotted. This is essentially the same curve as my 'cases per 100k tests' graph.

Unfortunately its only for England at the moment.
The 'About' also explains how it attempts to de-dup multiple tests and/or positives for the same person, I wonder if there is a different level of de-dup in the raw cases/positives numbers?

LimJim

2,274 posts

10 months

Monday 21st December 2020
quotequote all
Can I just double check you are using xlsx?

Terminator X

11,166 posts

172 months

Monday 21st December 2020
quotequote all
steveT350C said:
Good call

eta:

Quoting Spruce Goose from the cure thread,

“ Heart disease, cancer, etc. Deaths are surprisingly down this year. Cv19 deaths are up, people that die at home are up.

The average age of cv19 deaths is 82, 1 year younger than normal average death age.

Hospitals are overwhelmed, but surprisingly no different to last 20 years.”


Having the above’s data, and indeed all relevant data, packaged and presented in an easily shared and understood format has to be done.

I’ll build a PowerPoint presentation based on what gets posted here.

Edited by steveT350C on Sunday 20th December 15:01
It is frustrating albeit not surprising that simple facts and figures re so hard to find re Covid. If only either the Govt or the media (perish the thought) just presented all the adults with facts then let us risk assess for ourselves spin

TX.

coanda

2,296 posts

158 months

Monday 21st December 2020
quotequote all
smartypants said:
coanda said:
It's essential that all graphs posted can be recreated from source data by anyone with a basic level of spreadsheet knowledge, in order to build confidence in the data sets. Poorly presented graphs or graphs with no source data are not very useful and can discredit something that is otherwise genuine. We see this at every government presentation and throughout all industry.

Data manipulation is underhand and uncalled for, and in some case just dangerous. Let the data speak for itself.
Be lovely if HMG would abide by the same notion smile
I hold myself to higher standards than HMGs.

smartypants

50,075 posts

137 months

Monday 21st December 2020
quotequote all
I hope we all do.

Bill

45,524 posts

223 months

Monday 21st December 2020
quotequote all
bigandclever said:
steveT350C said:
Quoting Spruce Goose from the cure thread,

The average age of cv19 deaths is 82, 1 year younger than normal average death age.
I know we said ‘no arguments’, but the average Covid death age is 82.4; the average non-Covid death age is 81.5.

So average covid-death age is older, not younger.
That makes sense, very few under 50 have died of Covid.

Jevvy

220 posts

137 months

Monday 21st December 2020
quotequote all
This chap does some excellent work with the data and adds some observations / questions too.

https://twitter.com/RealJoelSmalley

A few of his recent charts copied below.


















Otispunkmeyer

11,233 posts

123 months

Monday 21st December 2020
quotequote all
Has anyone seen any US thanks giving data? Just seen anecdotes about how the prognosis was more more more death and despair if people meet up for Thanksgiving. But they did and the data suggests otherwise...

Elysium

Original Poster:

10,725 posts

155 months

Monday 21st December 2020
quotequote all
Otispunkmeyer said:
Has anyone seen any US thanks giving data? Just seen anecdotes about how the prognosis was more more more death and despair if people meet up for Thanksgiving. But they did and the data suggests otherwise...
US and EU27 fatalities per million from the datagraver website:


bigandclever

11,471 posts

206 months

Monday 21st December 2020
quotequote all
That’s not data, it’s a terrible, unintelligible chart.

Elysium

Original Poster:

10,725 posts

155 months

Monday 21st December 2020
quotequote all

coanda

2,296 posts

158 months

Monday 21st December 2020
quotequote all
bigandclever said:
That’s not data, it’s a terrible, unintelligible chart.
Let's see your attempt at answering Otis' question then.

bigandclever

11,471 posts

206 months

Monday 21st December 2020
quotequote all
coanda said:
bigandclever said:
That’s not data, it’s a terrible, unintelligible chart.
Let's see your attempt at answering Otis' question then.
It’s unreadable. Maybe it was great on the poster’s screen, but by the time it gets to mine it’s illegible.

eg


BlackTails

620 posts

23 months

Monday 21st December 2020
quotequote all
smartypants said:
Just data. No hypothesis.
Good idea. And yet the very next post contained the following hypotheses:

Elysium said:
This shows the vast scale of infections that was likely in the spring, but that we were completely unable to detect. It also shows that, despite massive increases in testing and case numbers we must still be missing a great many cases in the autumn.

Admissions and are deaths barely visible at this scale.
A data only thread is a good idea but unattainable. The temptation to provide a commentary (opinion, speculation or conclusion, take your pick) on the data is overwhelming and irresistible.

So, a suggestion: any interpretation - any - is prefaced by “in my opinion”, or “in my view”, or “I infer” or something similar. If you’re not sure whether what you’re saying is fact or your view of the facts, err on the side of caution.

Counting down 3-2-1 for this idea to be roundly ignored...