Discussion > Corona virus hockey stick
Neil Ferguson's track record
tomo
It is neither honest nor useful to compare the top end of a possible range if no mitigation measures were taken with reality. Any model that used the known factors at the time (R0 around 2.4, mortality around 1%) would have got to half a million over a few weeks. Latimer clearly lacks basic numeracy.
Ferguson's more relevant prediction of the outcome if PPE, social distancing and people staying home were deployed now seems to be low, I guess it assumed we would not ease restrictions so soon.
It is similarly dishonest to assess a conditional forecast - how many deaths if bird flu mutates into a lethal variant - when that didn't happen.
Where did the numbers for predicted deaths from foot and mouth disease originate? How does an animal disease kill 50,000 people?
More PFCN (Post First Check Later Never).
Phil Clarke
Where did the numbers for predicted deaths from foot and mouth disease originate? How does an animal disease kill 50,000 people?
More PFCN (Post First Check Later Never).
Oct 4, 2020 at 2:03 PM Phil Clarke
Are you claiming that Climate Science predictions remain the worst?
golf charlie
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019GL085378
Phil Clarke
Ferguson has a responsibility to challenge misinterpretations of his predictions & did he do that? - hell no.
I'm sure Ladbrokes would love to get him on board as horse racing pundit.
tomo
This was Twitter, not the Proceedings of the Royal Society.
I asked Alder where he got his FMD numbers - crickets.
Phil Clarke
Where did the numbers for predicted deaths from foot and mouth disease originate? How does an animal disease kill 50,000 people?
Oct 4, 2020 at 2:03 PM | Unregistered CommenterPhil Clarke
Phil - I think it should have been mad cow disease, not foot and mouth disease:
"In 2002, Ferguson predicted that up to 50,000 people would likely die from exposure to BSE (mad cow disease) in beef. "
Martin A
Yah, I wondered about that (along with Brid Flu (sic). The actual prediction ran:-
we estimate the 95% confidence interval for future vCJD mortality to be 50 to 50,000 human deaths considering exposure to bovine BSE alone
- Which naturally became 'Ferguson predicted 50,000 BSE deaths'. IOW - Its exactly the same misrepresentation.
I'd be the first to agree that an estimate that ranges over three orders of magnitude is not brilliant but the lower number assumes mitigation, which did occur. (Here's the paper). Same misrepresentation.
I wonder if Tomo feels a responsibility to challenge himself ;-)
Phil Clarke
https://twitter.com/latimeralder/status/1312833782496591875
Phil Clarke
I wonder if Tomo feels a responsibility to challenge himself ;-)
Oct 4, 2020 at 9:50 PM Phil Clarke
But you still believe in Gergis and Mann. Where is your Hockey Stick credibility?
golf charlie
Linking to that Tweet is an endorsement of sorts - it seems clear that Ferguson isn't prepared to either expose his (publicly funded) modelling or defend his estimates.
He's been right about everything and is unequivocally good value to the taxpayer then Phil?
yeah ... right-oh 50 to 50,000 at 95% confidence - what's 1000:1 spread in estimates compared to grant applications and consultancy billing? - most scientists I know would say "haven't a clue"
tomo
https://www.imperial.ac.uk/news/197875/codecheck-confirms-reproducibility-covid-19-model-results/
Phil Clarke
So bookmakers everywhere need to be afraid then.
Covidsim.org looks to be angling for dosh from the 3rd world (via WHO/UN/MRC grants?) - if you look at the list of places covered - many of which are simply incapable of providing sensible reporting of an infection.... question is - why are better data providers not listed?
It's all academic now innit? (50 of them at Imperial)
elsewhere:
Overheard at this morning's @RoyalStatSoc policy meeting:
"Do you think we should issue a statement on this Covid data screw-up?"
"To be honest I'm not sure you need a professor of statistics to tell you how many columns there are in an Excel spreadsheet"
tomo
Being pedantic, it was rows, the limit being 2^20 or 1,048,576.
I was often asked, usually by someone who fancied themselves as a 'power user' if we could do some part of the application I was designing in Excel with a few macros. My answer was always that it's a great little desktop tool but for the reason above, and because it is impossible to secure, and it has a tendency to mangle data formats, and it lacks any kind of audit capability and half a dozen other reasons it has absolutely no place in a production ETL system.
And don't get me started on Access databases.
Phil Clarke
Excel and Access?
pfffffff......
When you give bureaucrats chainsaws rather than Swiss Army knives - they generally can't work them.....
The Ofgem Renewables Database being a case in point
tomo
This smacks of departments tinkering with data for years without ever needing to get it right... though come to think of it the big software companies seem to do the same, they're just better at covering it up. Government IT has needed a shakeup for years.
TinyCO2
Government IT has needed a shakeup for years.
Test and Trace is mainly run by Serco.
Phil Clarke
Clarky
I feel confident in predicting an awards ceremony where certificates and maybe even medals (peerages even?) are handed out to civil servants who've supervised Serco.
tomo
More at The Register - which finally seems to be upping its game a bit.
The outsourcer (Serco) is not responsible for the design or overall management of NHS Test and Trace, nor its IT systems. Nonetheless the outsourcer managed to leak the email addresses of around 300 human contact tracers, the people paid to find out who an infected person was in touch with back in May.
tomo
Just after the company I worked for took over the main competitor I went to visit the factory to see if there was anything to learn to improve our factory. Their IT dept of course said they were wonderful and their MRP system had very few mods as everyone worked to the official system. I then went into the production control dept and found a few untouched MRP printouts on the managers desk and an army of people typing furiously at PC keyboards and running Excel macro's.
If you see a system run on an Excel spreadsheet go and ask the IT dept for their timescales on developing and delivering a program asked for by a user. The answer will tell you why Excel is so popular but also so dangerous.
Muppets are us
I've seen Excel lash-ups that were like the IT equivalent of Jenga
tomo
In fact, coronavirus is a serious problem. I have many acquaintances who have fallen ill and there are a number of deaths. People won't understand until they are confronted with it themselves.
adrianabell
clipe
Interesting the comparison between Mann's hockey stick and the UK govt's corona hockey stick - both presented with the intention of instilling fear.
Steve Mcintyre was right when he said you should be very wary when someone presents a hockey stick graph.
Note the huge jump in the first derivative at 15 September. "This graph, which is not a prediction....".
https://youtu.be/drTCzZCkqgw?t=283
[Sorry, I have forgotten how to make the URL clickable.]