https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

The Rev Dodgson said:

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

No woner my garden is looking a bit dry.

sibeen said:

The Rev Dodgson said:

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

No woner my garden is looking a bit dry.

Yeah, and your bloody green buildings are sucking all the rain down from Sydney too.

this is why i don’t sports

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

Extremely interesting. I already have qualms about statistics in medicine, where small numbers of people are justified by what I consider misuse of the statistical significance p-value. Particularly when added to to the Texas sharpshooter problem.

I’m also on record as saying that the relationship between cause and effect for physiotherapy is back to front. It’s not a case of “those who complete the schedule of exercises get well faster” but a case of “only those who are well can complete the schedule of exercises”.

> it’s not unusual for a treatment to be tested on fewer than 10 people. MBI, has no sound theoretical basis. MBI really does is lower the standard of evidence and increase the false positive rate.

Fewer than 10 people is useless.

> It is an amalgam of two statistical approaches — frequentist and Bayesian

As an old frequentics, I’m OK with that.

> “Assume that MBI has a 27% false positive rate and a 60% false negative rate”

Aagh! Crash and burn. I refuse to accept even the p-value false positive rate of 5%. Suppose there are 100 methods of which none are really useful, a common enough occurrence. The p-value gives a glowing report for 5% of these methods. I refuse to accept those 5% as valid treatments. A 27% false positive rate for MBI is totally useless. Especially with a 60% false negative rate.

2 sigma is 5% false positives and I won’t even look at the results.
2.5 sigma is the first point at which I’ll take notice. That’s 1.2% false positives.
3 sigma I like. That’s 0.3% false positives.

The “gold standard” in physics is 7 sigma.

Thank you for the warning, sibeen.

Thanks for sharing.

:-)

The Rev Dodgson said:

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

really?

In the densest part of the Dandenongs.. It could be 99%

mollwollfumble said:

The “gold standard” in physics is 7 sigma.

Thank you for the warning, sibeen.

But medicine isn’t physics.

It’s engineering, and if you looked for 7 sigma in engineering, you wouldn’t do anything.

roughbarked said:

The Rev Dodgson said:

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

really?

In the densest part of the Dandenongs.. It could be 99%

Meaning? Fuck all hits the soil.

roughbarked said:

The Rev Dodgson said:

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

really?

In the densest part of the Dandenongs.. It could be 99%

What could?

The Rev Dodgson said:

roughbarked said:

The Rev Dodgson said:

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

really?

In the densest part of the Dandenongs.. It could be 99%

What could?

What you said.

Trees suck most of it before it reaches the soil level.

roughbarked said:

roughbarked said:

The Rev Dodgson said:

If only it was limited to sports.

Yesterday I learned that a couple of trees in planters on the roof of a building can absorb 96% of rainfall. It wasn’t clear if that was 96% of the rain falling on that building or the whole of Melbourne.

really?

In the densest part of the Dandenongs.. It could be 99%

Meaning? Fuck all hits the soil.

What does that have to do with a couple of trees in planters on top of a building?

The Rev Dodgson said:

roughbarked said:

roughbarked said:

really?

In the densest part of the Dandenongs.. It could be 99%

Meaning? Fuck all hits the soil.

What does that have to do with a couple of trees in planters on top of a building?

It really is the same deal. They suck all that falls on them.

roughbarked said:

The Rev Dodgson said:

roughbarked said:

Meaning? Fuck all hits the soil.

What does that have to do with a couple of trees in planters on top of a building?

It really is the same deal. They suck all that falls on them.

FFS, why woould they risk sharing it with others not beneficial to their growth?

roughbarked said:

The Rev Dodgson said:

roughbarked said:

Meaning? Fuck all hits the soil.

What does that have to do with a couple of trees in planters on top of a building?

It really is the same deal. They suck all that falls on them.

So a couple of trees in planters, covering < 10% of the area of the building are not going to absorb 95% of the rain falling on the building, let alone reduce the water reaching the stormwater system by 95%, are they?

what about the 2 sigma of bloom

The Rev Dodgson said:

roughbarked said:

The Rev Dodgson said:

What does that have to do with a couple of trees in planters on top of a building?

It really is the same deal. They suck all that falls on them.

So a couple of trees in planters, covering < 10% of the area of the building are not going to absorb 95% of the rain falling on the building, let alone reduce the water reaching the stormwater system by 95%, are they?

No. they absorb only what they can catch and you knew that. Their roots will chase whatever runs downhill and you knew that.

The Rev Dodgson said:

mollwollfumble said:

The “gold standard” in physics is 7 sigma.

Thank you for the warning, sibeen.

But medicine isn’t physics.

It’s engineering, and if you looked for 7 sigma in engineering, you wouldn’t do anything.

But in engineering, we don’t work on a 5% chance of failure in each of, say, fifteen different failure modes. Engineers have been known to work with 2.5 sigma, a 1.2% chance of failure, but we don’t like to. 3 sigma for engineering is better, as I said above.

That just makes me think. Do we sometimes look for 7 sigma in engineering? Perhaps we do.

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

“Pressure vessels use a factor of safety of 3.5 to 4.0, automobiles use 3.0”. So for pressure vessels, if the calculation is correct, that’s actually well over 7 sigma.

mollwollfumble said:

The Rev Dodgson said:

mollwollfumble said:

The “gold standard” in physics is 7 sigma.

Thank you for the warning, sibeen.

But medicine isn’t physics.

It’s engineering, and if you looked for 7 sigma in engineering, you wouldn’t do anything.

But in engineering, we don’t work on a 5% chance of failure in each of, say, fifteen different failure modes. Engineers have been known to work with 2.5 sigma, a 1.2% chance of failure, but we don’t like to. 3 sigma for engineering is better, as I said above.

That just makes me think. Do we sometimes look for 7 sigma in engineering? Perhaps we do.

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

“Pressure vessels use a factor of safety of 3.5 to 4.0, automobiles use 3.0”. So for pressure vessels, if the calculation is correct, that’s actually well over 7 sigma.

and the mean equilibrium?

roughbarked said:

The Rev Dodgson said:

roughbarked said:

It really is the same deal. They suck all that falls on them.

So a couple of trees in planters, covering < 10% of the area of the building are not going to absorb 95% of the rain falling on the building, let alone reduce the water reaching the stormwater system by 95%, are they?

No. they absorb only what they can catch and you knew that. Their roots will chase whatever runs downhill and you knew that.

It seems you agree with my original point then.

The Rev Dodgson said:

roughbarked said:

The Rev Dodgson said:

So a couple of trees in planters, covering < 10% of the area of the building are not going to absorb 95% of the rain falling on the building, let alone reduce the water reaching the stormwater system by 95%, are they?

No. they absorb only what they can catch and you knew that. Their roots will chase whatever runs downhill and you knew that.

It seems you agree with my original point then.

when have I ever naysayed?

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

The Rev Dodgson said:

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Safer that we aren’t.

The Rev Dodgson said:

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Precisely :)

sibeen said:

The Rev Dodgson said:

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Precisely :)

Flying the kite in the wind, doen’t always work.

roughbarked said:

sibeen said:

The Rev Dodgson said:

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Precisely :)

Flying the kite in the wind, doen’t always work.

s?

The Rev Dodgson said:

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Or think of it this way. There are what, 100,000 or so engineered buildings in the world.

A 27% incidence of complete structural failure (MBI statistics) is unacceptable.
A 5% incidence of complete structural failure (p-values) is unacceptable.
A 1.2% incidence of complete structural failure (2.5 sigma) is unacceptable.
A 0.3% incidence of complete structural failure (3 sigma) gives 300 or so complete structural failures of engineered buildings around the world.

As I said above, 3 sigma is what I aim for. But is that safe enough? You decide. What do you design to? 3.5 sigma would be better, 0.12% incidence of complete structural failure, 120 buildings world wide.

I don’t say that 7 sigma is what engineers should aim for, quite the reverse. 7 sigma is horribly expensive. I am saying that in some cases, particularly pressure vessels, some engineers will be effectively designing to that.

mollwollfumble said:

The Rev Dodgson said:

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Or think of it this way. There are what, 100,000 or so engineered buildings in the world.

A 27% incidence of complete structural failure (MBI statistics) is unacceptable.
A 5% incidence of complete structural failure (p-values) is unacceptable.
A 1.2% incidence of complete structural failure (2.5 sigma) is unacceptable.
A 0.3% incidence of complete structural failure (3 sigma) gives 300 or so complete structural failures of engineered buildings around the world.

As I said above, 3 sigma is what I aim for. But is that safe enough? You decide. What do you design to? 3.5 sigma would be better, 0.12% incidence of complete structural failure, 120 buildings world wide.

I don’t say that 7 sigma is what engineers should aim for, quite the reverse. 7 sigma is horribly expensive. I am saying that in some cases, particularly pressure vessels, some engineers will be effectively designing to that.

100,000? There must be way more than that.

But the point is, calculating a probability of failure assuming a normal distribution of defined input completely misses the things that cause nearly all structural collapses, which are:

1. Gross errors in construction
2. Gross errors in design
3. Stuff happening that was not considered in the design

or most likely, a combination of all 3.

Increasing the target sigma level does little if anything to reduce these problems.

mollwollfumble said:

The Rev Dodgson said:

mollwollfumble said:

Checking wikipedia: “Buildings commonly use a factor of safety of 2.0 for each structural member”. If those are steel members, and the calculation is accurate, then for dead loads we’re running pretty close to 7 sigma.

But since the calculation is not accurate, and the estimate of applied loads is even less accurate, we are nowhere near 7 sigma.

Or think of it this way. There are what, 100,000 or so engineered buildings in the world.

A 27% incidence of complete structural failure (MBI statistics) is unacceptable.
A 5% incidence of complete structural failure (p-values) is unacceptable.
A 1.2% incidence of complete structural failure (2.5 sigma) is unacceptable.
A 0.3% incidence of complete structural failure (3 sigma) gives 300 or so complete structural failures of engineered buildings around the world.

As I said above, 3 sigma is what I aim for. But is that safe enough? You decide. What do you design to? 3.5 sigma would be better, 0.12% incidence of complete structural failure, 120 buildings world wide.

I don’t say that 7 sigma is what engineers should aim for, quite the reverse. 7 sigma is horribly expensive. I am saying that in some cases, particularly pressure vessels, some engineers will be effectively designing to that.

“There are what, 100,000 or so engineered buildings in the world.”

What?

I had to get my extensions here engineered. The original building was engineered (and I have all that work). Gympie Regional Council requires engineering for all structures, and has done since before 2003.

I’ve got an idea.

How about Rev Dodgson send me specs on the last structure he designed and I’ll calculate the probability of failure.

mollwollfumble said:

I’ve got an idea.

How about Rev Dodgson send me specs on the last structure he designed and I’ll calculate the probability of failure.

The last structure I completed a design for is a good illustration of the futility of that process.

It was a crash barrier.

The Rev Dodgson said:

mollwollfumble said:

I’ve got an idea.

How about Rev Dodgson send me specs on the last structure he designed and I’ll calculate the probability of failure.

The last structure I completed a design for is a good illustration of the futility of that process.

It was a crash barrier.

LOL. I love it.

I’ve been concerned about all the new wire rope crash barriers that are going up in hundreds of locations in rural Victoria. Because when I looked up the Uni NSW tests I found that a car impacting the crash barrier during tests overturned with what would have been fatal injuries. In addition, the wire rope barriers have a bad habit of deflecting vehicles back into the stream of traffic potentially creating more accidents. The W-beam barriers are safer.

I did some calculations and found that the wire rope crash barriers were safest if the support posts are not concreted in. Concreting the posts in increased the deceleration force making it less safe. In addition, above a certain speed the wire rope snaps leading to loss of effectiveness and high fence repair costs. The ideal is to have the posts held just in compacted soil so they give way progressively through the impact so the impact force is spread out over a longer more uniform deceleration. And the force is taken elastically by the wire, minimising barrier repair costs because the wire and most of the posts are dislodged but undamaged.

Another issue I have with the installation of all these wire rope crash barriers is the narrowing of the emergency lane. Before the barrier there was no limit on completely opening the passenger side door, but with the addition of crash barriers this becomes impossible. Opening the passenger side door immediately is essential for nearly every type of medical emergency requiring a pull-over, as well as in the case of a car fire.

So, what vehicle did you test your crash barrier with? At what speed? At what angle? How long a crash barrier? Did it pass the test?

mollwollfumble said:

The Rev Dodgson said:

mollwollfumble said:

I’ve got an idea.

How about Rev Dodgson send me specs on the last structure he designed and I’ll calculate the probability of failure.

The last structure I completed a design for is a good illustration of the futility of that process.

It was a crash barrier.

LOL. I love it.

I’ve been concerned about all the new wire rope crash barriers that are going up in hundreds of locations in rural Victoria. Because when I looked up the Uni NSW tests I found that a car impacting the crash barrier during tests overturned with what would have been fatal injuries. In addition, the wire rope barriers have a bad habit of deflecting vehicles back into the stream of traffic potentially creating more accidents. The W-beam barriers are safer.

I did some calculations and found that the wire rope crash barriers were safest if the support posts are not concreted in. Concreting the posts in increased the deceleration force making it less safe. In addition, above a certain speed the wire rope snaps leading to loss of effectiveness and high fence repair costs. The ideal is to have the posts held just in compacted soil so they give way progressively through the impact so the impact force is spread out over a longer more uniform deceleration. And the force is taken elastically by the wire, minimising barrier repair costs because the wire and most of the posts are dislodged but undamaged.

Another issue I have with the installation of all these wire rope crash barriers is the narrowing of the emergency lane. Before the barrier there was no limit on completely opening the passenger side door, but with the addition of crash barriers this becomes impossible. Opening the passenger side door immediately is essential for nearly every type of medical emergency requiring a pull-over, as well as in the case of a car fire.

So, what did you test your crash barrier with? At what speed? At what angle? How long a crash barrier? Did it pass the test?

I just followed the Bridge Design Code, which has different static loads for different classes of containment. How this load relates to the actual dynamic loads under crash conditions, I have no idea,

We don’t see t

mollwollfumble said:

The Rev Dodgson said:

mollwollfumble said:

I’ve got an idea.

How about Rev Dodgson send me specs on the last structure he designed and I’ll calculate the probability of failure.

The last structure I completed a design for is a good illustration of the futility of that process.

It was a crash barrier.

LOL. I love it.

I’ve been concerned about all the new wire rope crash barriers that are going up in hundreds of locations in rural Victoria. Because when I looked up the Uni NSW tests I found that a car impacting the crash barrier during tests overturned with what would have been fatal injuries. In addition, the wire rope barriers have a bad habit of deflecting vehicles back into the stream of traffic potentially creating more accidents. The W-beam barriers are safer.

I did some calculations and found that the wire rope crash barriers were safest if the support posts are not concreted in. Concreting the posts in increased the deceleration force making it less safe. In addition, above a certain speed the wire rope snaps leading to loss of effectiveness and high fence repair costs. The ideal is to have the posts held just in compacted soil so they give way progressively through the impact so the impact force is spread out over a longer more uniform deceleration. And the force is taken elastically by the wire, minimising barrier repair costs because the wire and most of the posts are dislodged but undamaged.

Another issue I have with the installation of all these wire rope crash barriers is the narrowing of the emergency lane. Before the barrier there was no limit on completely opening the passenger side door, but with the addition of crash barriers this becomes impossible. Opening the passenger side door immediately is essential for nearly every type of medical emergency requiring a pull-over, as well as in the case of a car fire.

So, what did you test your crash barrier with? At what speed? At what angle? How long a crash barrier? Did it pass the test?

I just followed the Bridge Design Code, which has different static loads for different classes of containment. How this load relates to the actual dynamic loads under crash conditions, I have no idea,

We don’t see t

The Rev Dodgson said:

Bloody computers.

We don’t see the wire rope barriers much in NSW. I’ll have to check what RMS say about them.

The Rev Dodgson said:

The Rev Dodgson said:

I just followed the Bridge Design Code, which has different static loads for different classes of containment. How this load relates to the actual dynamic loads under crash conditions, I have no idea.

We don’t see the wire rope barriers much in NSW. I’ll have to check what RMS say about them.

There are wire rope barriers on the NSW Hume Highway. But strangely a different number of wires to the Vic ones.

Is there a guide to the bridge code (like there used to be for the wind loading code) that explains where all the numbers and equations come from?

The wind loading code is the only code that I have a lot of experience with, though I worked with people who were writing the BCAider software for the building code.

mollwollfumble said:

sibeen said:

https://fivethirtyeight.com/features/how-shoddy-statistics-found-a-home-in-sports-research/

May be of interest to some.

Extremely interesting. I already have qualms about statistics in medicine, where small numbers of people are justified by what I consider misuse of the statistical significance p-value. Particularly when added to to the Texas sharpshooter problem.

I’m also on record as saying that the relationship between cause and effect for physiotherapy is back to front. It’s not a case of “those who complete the schedule of exercises get well faster” but a case of “only those who are well can complete the schedule of exercises”.

> it’s not unusual for a treatment to be tested on fewer than 10 people. MBI, has no sound theoretical basis. MBI really does is lower the standard of evidence and increase the false positive rate.

Fewer than 10 people is useless.

> It is an amalgam of two statistical approaches — frequentist and Bayesian

As an old frequentics, I’m OK with that.

> “Assume that MBI has a 27% false positive rate and a 60% false negative rate”

Aagh! Crash and burn. I refuse to accept even the p-value false positive rate of 5%. Suppose there are 100 methods of which none are really useful, a common enough occurrence. The p-value gives a glowing report for 5% of these methods. I refuse to accept those 5% as valid treatments. A 27% false positive rate for MBI is totally useless. Especially with a 60% false negative rate.

2 sigma is 5% false positives and I won’t even look at the results.
2.5 sigma is the first point at which I’ll take notice. That’s 1.2% false positives.
3 sigma I like. That’s 0.3% false positives.

The “gold standard” in physics is 7 sigma.

Thank you for the warning, sibeen.

Always amusing when physicists try biology…

poikilotherm said:

Always amusing when physicists try biology…

Even more amusing when biologists try statistics…

mollwollfumble said:

poikilotherm said:

Always amusing when physicists try biology…

Even more amusing when biologists try statistics…

Usually have a pretty good grasp of how to apply stats to their field, then again, according to your thinking, every living thing is as simple as a bridge or building…

poikilotherm said:

mollwollfumble said:

poikilotherm said:

Always amusing when physicists try biology…

Even more amusing when biologists try statistics…

Usually have a pretty good grasp of how to apply stats to their field, then again, according to your thinking, every living thing is as simple as a bridge or building…

I should probably let mollwoll speak for himself, but I’d say his thinking was that living things are just as complicated as a bridge or a building.

Perhaps sometimes even more so.