Expert modelling

[DaveGrohl]

I wasn't quite sure where to put this so thought what the heck. You get some lovely little snippets of info in Money Week, this is from 22 May issue.

David Richards of British data company Wandisco, on the computer modelling underpinning Prof Neil Ferguson's predictions of the spread of Covid-19, which prompted the govt to put the country in lockdown, quoted in the Sun (of all places!):

"(A) buggy mess that looks more like a bowl of angel hair pasta than a finely tuned piece of programming..... We would fire anyone for developing code like this and any business that relied on it to produce software for sale would likely go bust."

 

[Smith31]

The past few weeks suggest the experts know very little, maybe time for some p45's

 

[DaveGrohl]

He's already got his of course, sadly not for any intellectual reasons though.

Also, it depends which expert you ask to begin with. Ancel Keys is one one that defo shouldn't have been listened to. I wonder how many early deaths he caused and is still causing?

 

[holwellcourtfarm]

I wasn't quite sure where to put this so thought what the heck. You get some lovely little snippets of info in Money Week, this is from 22 May issue.

David Richards of British data company Wandisco, on the computer modelling underpinning Prof Neil Ferguson's predictions of the spread of Covid-19, which prompted the govt to put the country in lockdown, quoted in the Sun (of all places!):

"(A) buggy mess that looks more like a bowl of angel hair pasta than a finely tuned piece of programming..... We would fire anyone for developing code like this and any business that relied on it to produce software for sale would likely go bust."

ALL modelling is a compromise. To be really accurate it needs truly comprehensive data sets, a complete understanding of the system being modelled and a range of fully documented "calibration events". Without these the margins of error increase massively.

Most computer models are a "best guess" at best, especially if the subject is any sort of natural system.

(I spent years at the EA "ground truthing" river flood models. All were inaccurate and some were dangerously so).

 

[yellowbelly]

Computer modelling and it's proponents should be outlawed.

Just look at their track record - BSE, foot and mouth and now Covid-19.

All done by self appointed "experts", who are just glorified guessers - and not very good ones at that.

 

[arcobob]

ALL modelling is a compromise. To be really accurate it needs truly comprehensive data sets, a complete understanding of the system being modelled and a range of fully documented "calibration events". Without these the margins of error increase massively.

Most computer models are a "best guess" at best, especially if the subject is Amy sort of natural system.

(I spent years at the EA "ground truthing" river flood models. All were inaccurate and some were dangerously so).

Have the EA advanced any since then?

 

[Exfarmer]

There is nothing whatsoever wrong with computer modelling, it is just the old adage Rubbish in, Rubbish out.
It will only ever Be as good as the information you give it. Surprisingly it is not that difficult and relies on algorithms, which underpin the whole of the internet.
it is not the programme, but the judgement on the information you are feeding it which matters most.
If you know your way round excel , you can soon build modelling programmes yourself and many do this in business judgements every day.
When viewing comments about Neil Ferguson and others in his field, it is also wise to be aware, of the deep seated animosity, possibly hatred between the modellers from Oxford and those from Imperial. This relates to a huge bust up some years ago when one professor accused another professor of only being in the position she was a in, by sleeping with her boss!
This lead to a long court battle and half of the modelling team leaving for Imperial.

 

[holwellcourtfarm]

Have the EA advanced any since then?

The flood model of the River Stort cost many tens of thousands and each run took 2 days of computer time. They couldn't get it to "calibrate" (reproduce known past flood events when fed the inputs known to have caused the flood). I looked through their "node diagram" which detailed how they thought the catchment fitted together and found they'd missed off a weir and the river loop it fed and got the relative level of 3 weirs wrong.

Since I left so have all the other staff who knew the local catchments well enough to do that node check.

You decide!

 

[holwellcourtfarm]

There is nothing whatsoever wrong with computer modelling, it is just the old adage Rubbish in, Rubbish out.
It will only ever Be as good as the information you give it. Surprisingly it is not that difficult and relies on algorithms, which underpin the whole of the internet.
it is not the programme, but the judgement on the information you are feeding it which matters most.
If you know your way round excel , you can soon build modelling programmes yourself and many do this in business judgements every day.
When viewing comments about Neil Ferguson and others in his field, it is also wise to be aware, of the deep seated animosity, possibly hatred between the modellers from Oxford and those from Imperial. This relates to a huge bust up some years ago when one professor accused another professor of only being in the position she was a in, by sleeping with her boss!
This lead to a long court battle and half of the modelling team leaving for Imperial.

Modelling simple systems is indeed straightforward. Modelling complex ones is nearly impossible to do with any real accuracy.

Look at all the economic models which failed to predict the 2008 crash. The modelers all said afterwards that they had excluded extreme scenarios as "Black Swan events" because running them made their models unstable.

It's the extreme events that REALLY matter for resilience planning!

 

[arcobob]

The flood model of the River Stort cost many tens of thousands and each run took 2 days of computer time. They couldn't get it to "calibrate" (reproduce known past flood events when fed the inputs known to have caused the flood). I looked through their "node diagram" which detailed how they thought the catchment fitted together and found they'd missed off a weir and the river loop it fed and got the relative level of 3 weirs wrong.

Since I left so have all the other staff who knew the local catchments well enough to do that node check.

You decide!

That sounds about right but the sad thing is important decisions are still being made based on this flawed data. A more disfunctional government quango I never did see.

 

[DaveGrohl]

Modelling simple systems is indeed straightforward. Modelling complex ones is nearly impossible to do with any real accuracy.

Look at all the economic models which failed to predict the 2008 crash. The modelers all said afterwards that they had excluded extreme scenarios as "Black Swan events" because running them made their models unstable.

It's the extreme events that REALLY matter for resilience planning!

They weren't really economic models though were they, just predictions it would all be ok or better. Lots of people knew in advance something bad was coming, just not quite how it would play out. It still hasn't played out yet.

 

[DaveGrohl]

There is nothing whatsoever wrong with computer modelling, it is just the old adage Rubbish in, Rubbish out.
It will only ever Be as good as the information you give it. Surprisingly it is not that difficult and relies on algorithms, which underpin the whole of the internet.
it is not the programme, but the judgement on the information you are feeding it which matters most.
If you know your way round excel , you can soon build modelling programmes yourself and many do this in business judgements every day.
When viewing comments about Neil Ferguson and others in his field, it is also wise to be aware, of the deep seated animosity, possibly hatred between the modellers from Oxford and those from Imperial. This relates to a huge bust up some years ago when one professor accused another professor of only being in the position she was a in, by sleeping with her boss!
This lead to a long court battle and half of the modelling team leaving for Imperial.

There was an article in The Times a few weeks ago about precisely this, made for interesting reading.

 

[arcobob]

There is nothing whatsoever wrong with computer modelling, it is just the old adage Rubbish in, Rubbish out.
It will only ever Be as good as the information you give it. Surprisingly it is not that difficult and relies on algorithms, which underpin the whole of the internet.
it is not the programme, but the judgement on the information you are feeding it which matters most.
If you know your way round excel , you can soon build modelling programmes yourself and many do this in business judgements every day.
When viewing comments about Neil Ferguson and others in his field, it is also wise to be aware, of the deep seated animosity, possibly hatred between the modellers from Oxford and those from Imperial. This relates to a huge bust up some years ago when one professor accused another professor of only being in the position she was a in, by sleeping with her boss!
This lead to a long court battle and half of the modelling team leaving for Imperial.

Was she good at it?

 

[Agrivator]

We used to do modelling at primary school. Plasticine and pigs come to mind.

 

[JP1]

I wasn't quite sure where to put this so thought what the heck. You get some lovely little snippets of info in Money Week, this is from 22 May issue.

David Richards of British data company Wandisco, on the computer modelling underpinning Prof Neil Ferguson's predictions of the spread of Covid-19, which prompted the govt to put the country in lockdown, quoted in the Sun (of all places!):

"(A) buggy mess that looks more like a bowl of angel hair pasta than a finely tuned piece of programming..... We would fire anyone for developing code like this and any business that relied on it to produce software for sale would likely go bust."

I only came in this thread to see if it was a way of @Bald Rick getting two oooh threads running as a way of increasing the workload to become Honorary Mod ........