You're right, but I'm not doing so disingenuously. I don't believe you
can separate the two: If you're going to perform any risk
calculations, you have to sooner or later calculate an
expectation. The probabilities you'll put into that will depend on
collected data, and will be quite unreliable if the events being
assessed are rare, which the interesting ones often are. I'm
suggesting it's (numerically speaking) better to lump risks together
than try to model individual causes too finely.
Yes! You add in everything that the data you have represents. If your data is deaths, then you don't make a distinction between deaths. You can only make a distinction when the data makes that distinction for you (ie you're not imputing a classification yourself, after the data is collected). Anything else is mixing assumptions with facts in ways that are hard to unravel, if not by you, then certainly by others.
No, you're not computing full failure probabilities by ignoring the terrorist risk. What you're computing in your example is conditional probability, ie probability of a failure conditional on a nonterrorist event. That's not a probability, and can't be used to make decisions on its own. At best, it may be an intermediate result, but if you're doing that, you've now got two sources of modelling error, the error in the conditional probability and the error in the estimate of the probability of a terrorist event that is needed to complete the calculation.
You don't fairly lump in the risk of stroke due to meth abuse in air travel,
even though a passenger may indeed suffer that death while flying.
You do. It's still a death in flight, and will show up in casualty statistics(*).
If you deliberately ignore that risk, your safety probabilities (which are really conditionals) will be too rosy, and you won't be able to reconcile them with other data. And if your calculations get passed to someone else who uses them in another calculation, they may well not realize that the omission skews their own results as well.
(*) Even if it seems obvious that it shouldn't in this case, an actual determination of the cause of death might be delayed, and not taken into account when the statistics are compiled.
The two types of error aren't equally important. Lumping failure modes together leads to worst case analysis, which is safer than best case analysis. You're worried that people will be overconfident from passing a worst case analysis? As opposed to what? Being reasonably confident from passing a best case analysis instead? (Getting things exactly right isn't realistic, do you think otherwise?). I'd argue that confidence is less important than actual physical measures that are ready when needed.
For the case at hand, limiting population density near the runway will not in
any way address the terrorist case. They will go wherever the density is
anyway. It would address the case of short landings and failures on takeoff.
The question is, just how often do those happen a mile away.
Well, at around 150mph it would take about 25 seconds to reach a mile.
If the airplane takes off and has a fatal failure around 12 seconds into the climb, then by symmetry a rough estimate is that the crash will occur a mile
from the takeoff point. This page has data on accidents during the initial climb phase. Clearly this is not a negligible probability, even though it is aggregated over many airports.
Are you saying that "allocating according to preferences" uses a global utility function? Because the whole point of
the cake cutting problem is that each participant has different utility functions.
No, I'm saying fairness is a global utility function. In the cake cutting problem, the goal is to maximize the minimum sized piece that someone might get. If you generalize, that's when you introduce utilities that can be different for each player, etc.
I don't think so. Cake slicing problems (aka dividing fairly) have nothing to do with efficiency (aka allocating according to preferences). One has a global utility function, and the other has multiple individual utility functions.
As Bell is Australian, did he have the right to operate a business in the US on his visa? This is one of those questionable issues facing foreign startups a lot. If he couldn't legally operate a business in the US, who's responsible for and owns the video footage? I assume Steele considers herself the American partner who actually has all those rights?
IANAL and all that.
Your reply illustrates the classic problem when estimating failure
probabilities: Instead of being inclusive, you try to fix a narrow
definition that represents a failure event, which causes your final
probability estimates to be too low. This also happens with nuclear disasters and black swan type market crashes, etc.
I would suggest that your original question "When was the last time we heard of a fighter jet
crashing into an apartment complex?" is badly posed. It's too specific to be truly useful, and if answered precisely, leads to an overly optimistic conclusion.
All airplane crashes are different, of
course, but they are still crashes. I've been to airshows myself, and
I've seen military aircraft (the F15 flying straight up like a candle
is most impressive, btw), and incidentally I've also witnessed a crash into
a crowd at such an event.
But is it "landing" if you never hit the ground? Also, I can see this being a
problem for aircraft carrier jargon, although I seem to recall someone
saying that landing
on a carrier is like a "controlled crash".
Would you be happier if the US became a hereditary monarchy? You could have the same guy in power for about 50 years, then he gets replaced by his favourite snot-nosed kid.
You're very adorable. The movie and music industries don't age, sleep, or
need to do any work to get money. We will inevitably have to do all three.
Do you always compare apples and oranges? "The music industry" doesn't age or sleep, and neither does "The public". "The music executives" age, sleep, and work, just like "The members of the public".
Unlike patent trolls, CSIRO actually creates technology through research.
They deserve payment, unlike most who file their claims through East Texas.
No, I'm sorry, I think they do not.
As much as I admire the CSIRO for the great quality of research it is doing, I don't accept patents as solving a problem that must be solved, and I therefore don't accept that when it's some good guys doing the patenting, they should get money just because they're good.
Patents have serious issues which aren't necessary to be rehashed, I just want to argue against the phrase "they deserve payment" here. This makes it sound like the scientists aren't being paid, and are about to go begging in the streets because some US corp isn't paying its bill. It's hyperbolic.
The scientists have already been paid. The CSIRO employs scientists and pays them an annual salary to produce research. The CSIRO isn't some market speculator who does this just to make obscene amounts of money, it's a government body whose mission is scientific research to benefit the Australian people and humanity at large. So the usual (weak) arguments that the research wouldn't happen without the prospect of huge wealth don't apply at all here.
Secondly, what do you think happens when the patent licenses are paid? Where does the money go? It doesn't go only to the department who invented the WiFi algorithms. It gets distributed throughout the CSIRO. So you'll get scientists working on rabbit control getting funded by the WiFi patents. Now, this is great - extra money for scientists, etc - but it makes no sense whatsoever from a patent perspective, if you believe in them.
WiFi patents (for those who accept the rationale) are there to promote research and justify expenditure in WiFi technology. There's no question of promoting rabbit research. In fact, doing so is a complete perversion of the patent concept, as the technology which earned the patent licenses gets penalized (it doesn't receive all the money it should) in favour of unrelated technology which hasn't anything to do with the patent.
I'm opposed to patents, I think they're wrong, and I make no exception when the good guys receive the license money.
Which is why we have to keep following and mentioning it on slashdot, even if we already know this. We (the public) can outlast them, if a relatively small number of us are dedicated to bringing the attempts to light, again and again, for as long as it takes.
They'll just pick the closest person that they think
fits the person that they saw earlier. And earlier might be hours, days, or
even weeks or longer. Academic studies have shown that if you give the
average person a lineup of random innocent people that most people will
finger one of them for the crime.
And that, kids, is why you don't wear a beard or keep long Jesus hair.
Remember, the Son Of God Himself was innocent, but with the way he looked he still got fingered in a lineup and ended up on the cross. Hippies, eh? They never learn!
But let's assume
he's in the continental US so we've got something to talk about (and just to
piss off the whiners who complain every time this is assumed).
Agreed. Let's also assume he lives in Minnesota, because he mentioned the alpine countryside, which suggests he's got Norwegian ancestry from the great mountains of Central Europe.
First: tulips were grown in increasing quantities.
Tulips are perishable goods, with a very long growing period, and a low rate of reproduction. They were considered valuable precisely because not everybody could grow them at will. That's why your analogy with printing money breaks down.
Second: people were trading options (yes, derivatives, oh my) on the futures
market, and then government came out with a law that allowed people to drop
contracts and only be liable for 3.5% of the value. You don't see how that
can create a problem that later can be blamed on 'free market', where in fact
there was no such thing again?
If the law considers a contract partially or fully unenforceable, then anyone entering into such a contract nevertheless is ignorant and a sucker. Why are we expected to believe that markets of such people are good at predicting things? Are they secret masochists?
By the way, 80 years of war, I always was wondering who in their right mind
would fight wars for 80 years without government actively promoting them.
Off topic, but quite simple really. In the 17th century, most soldiers were mercenaries, whose motivation was money and pillage. Fighting wars was a career choice, basically, and soldiers frequently switched sides whenever their current employer was defeated, lost interest or didn't pay enough. It's best to think of them as criminal gangs terrorizing large parts of Europe.
Tulips are grown at will, they are nothing more than Federal reserve notes
without the Federal reserve. They ARE paper.
Is that so? You are saying anybody could grow valuable tulips in the 17th century? How'd they get the seeds and offsets? How'd they control the patterns? How'd they speed up the 10 year growing time?
You don't need to have a central monetary authority to have a bubble based on
creation of fake currency if fake currency can be created by anybody anyway.
It's perfect, it's as if everybody had their own printing press.
Interesting, you're blaming market participants for the failures of markets. How does that advance the claim that markets are good at prediction? Would it help if markets didn't allow people to join? Maybe someone should do a study on that.
On a grander scale, there is, for example, the issue of the data
obtained by the Nazi's WW2 involuntary medical experiments and whether it
should be used or not. This is a test of personal ethics, not sanity.
I have to agree with the OP, data is data no matter where it came from. HOWEVER, there is still no point in using the WW2 medical data, for reasons which do not require ethics to explain.
Science depends on repeatability at will.
One time data sets can be historically interesting, but if they can't be duplicated today and tomorrow, then they don't strictly add to our understanding of the world. This is the problem with the WWII Nazi experiments. They aren't acceptable and are incompatible with human rights, therefore they can't be duplicated and refined/improved. They are best ignored and left for the historians and philosophers.
Historically speaking, paying for music was common for rich people, though.
There were always minstrels in castles, paid with money or otherwise.
And musical performances in theatres were for pay, too.
I suspect that the big boost in music spending had more to do with the relative affluence of the 60s generation, and their kids in the 80s.
Slashdot Mirror
You're right, but I'm not doing so disingenuously. I don't believe you can separate the two: If you're going to perform any risk calculations, you have to sooner or later calculate an expectation. The probabilities you'll put into that will depend on collected data, and will be quite unreliable if the events being assessed are rare, which the interesting ones often are. I'm suggesting it's (numerically speaking) better to lump risks together than try to model individual causes too finely.
Yes! You add in everything that the data you have represents. If your data is deaths, then you don't make a distinction between deaths. You can only make a distinction when the data makes that distinction for you (ie you're not imputing a classification yourself, after the data is collected). Anything else is mixing assumptions with facts in ways that are hard to unravel, if not by you, then certainly by others.
You do. It's still a death in flight, and will show up in casualty statistics(*). If you deliberately ignore that risk, your safety probabilities (which are really conditionals) will be too rosy, and you won't be able to reconcile them with other data. And if your calculations get passed to someone else who uses them in another calculation, they may well not realize that the omission skews their own results as well.
(*) Even if it seems obvious that it shouldn't in this case, an actual determination of the cause of death might be delayed, and not taken into account when the statistics are compiled.
Well, at around 150mph it would take about 25 seconds to reach a mile. If the airplane takes off and has a fatal failure around 12 seconds into the climb, then by symmetry a rough estimate is that the crash will occur a mile from the takeoff point. This page has data on accidents during the initial climb phase. Clearly this is not a negligible probability, even though it is aggregated over many airports.
No, I'm saying fairness is a global utility function. In the cake cutting problem, the goal is to maximize the minimum sized piece that someone might get. If you generalize, that's when you introduce utilities that can be different for each player, etc.
Mrs Slocombe, is that you?
I don't think so. Cake slicing problems (aka dividing fairly) have nothing to do with efficiency (aka allocating according to preferences). One has a global utility function, and the other has multiple individual utility functions.
As Bell is Australian, did he have the right to operate a business in the US on his visa? This is one of those questionable issues facing foreign startups a lot. If he couldn't legally operate a business in the US, who's responsible for and owns the video footage? I assume Steele considers herself the American partner who actually has all those rights? IANAL and all that.
I would suggest that your original question "When was the last time we heard of a fighter jet crashing into an apartment complex?" is badly posed. It's too specific to be truly useful, and if answered precisely, leads to an overly optimistic conclusion.
All airplane crashes are different, of course, but they are still crashes. I've been to airshows myself, and I've seen military aircraft (the F15 flying straight up like a candle is most impressive, btw), and incidentally I've also witnessed a crash into a crowd at such an event.
But is it "landing" if you never hit the ground? Also, I can see this being a problem for aircraft carrier jargon, although I seem to recall someone saying that landing on a carrier is like a "controlled crash".
About 11 years ago, on September 11 2001? Well, it wasn't an apartment complex as such, but it was definitely a fighter (commandeered) jet.
Pick your doom.
Do you always compare apples and oranges? "The music industry" doesn't age or sleep, and neither does "The public". "The music executives" age, sleep, and work, just like "The members of the public".
No, I'm sorry, I think they do not.
As much as I admire the CSIRO for the great quality of research it is doing, I don't accept patents as solving a problem that must be solved, and I therefore don't accept that when it's some good guys doing the patenting, they should get money just because they're good.
Patents have serious issues which aren't necessary to be rehashed, I just want to argue against the phrase "they deserve payment" here. This makes it sound like the scientists aren't being paid, and are about to go begging in the streets because some US corp isn't paying its bill. It's hyperbolic.
The scientists have already been paid. The CSIRO employs scientists and pays them an annual salary to produce research. The CSIRO isn't some market speculator who does this just to make obscene amounts of money, it's a government body whose mission is scientific research to benefit the Australian people and humanity at large. So the usual (weak) arguments that the research wouldn't happen without the prospect of huge wealth don't apply at all here.
Secondly, what do you think happens when the patent licenses are paid? Where does the money go? It doesn't go only to the department who invented the WiFi algorithms. It gets distributed throughout the CSIRO. So you'll get scientists working on rabbit control getting funded by the WiFi patents. Now, this is great - extra money for scientists, etc - but it makes no sense whatsoever from a patent perspective, if you believe in them.
WiFi patents (for those who accept the rationale) are there to promote research and justify expenditure in WiFi technology. There's no question of promoting rabbit research. In fact, doing so is a complete perversion of the patent concept, as the technology which earned the patent licenses gets penalized (it doesn't receive all the money it should) in favour of unrelated technology which hasn't anything to do with the patent.
I'm opposed to patents, I think they're wrong, and I make no exception when the good guys receive the license money.
Which is why we have to keep following and mentioning it on slashdot, even if we already know this. We (the public) can outlast them, if a relatively small number of us are dedicated to bringing the attempts to light, again and again, for as long as it takes.
I heard about that! Snakes on a plain...
And that, kids, is why you don't wear a beard or keep long Jesus hair. Remember, the Son Of God Himself was innocent, but with the way he looked he still got fingered in a lineup and ended up on the cross. Hippies, eh? They never learn!
Depends if they're the Yahoo's from New Zealand!
Agreed. Let's also assume he lives in Minnesota, because he mentioned the alpine countryside, which suggests he's got Norwegian ancestry from the great mountains of Central Europe.
Tulips are perishable goods, with a very long growing period, and a low rate of reproduction. They were considered valuable precisely because not everybody could grow them at will. That's why your analogy with printing money breaks down.
If the law considers a contract partially or fully unenforceable, then anyone entering into such a contract nevertheless is ignorant and a sucker. Why are we expected to believe that markets of such people are good at predicting things? Are they secret masochists?
Off topic, but quite simple really. In the 17th century, most soldiers were mercenaries, whose motivation was money and pillage. Fighting wars was a career choice, basically, and soldiers frequently switched sides whenever their current employer was defeated, lost interest or didn't pay enough. It's best to think of them as criminal gangs terrorizing large parts of Europe.
Is that so? You are saying anybody could grow valuable tulips in the 17th century? How'd they get the seeds and offsets? How'd they control the patterns? How'd they speed up the 10 year growing time?
Interesting, you're blaming market participants for the failures of markets. How does that advance the claim that markets are good at prediction? Would it help if markets didn't allow people to join? Maybe someone should do a study on that.
The first known bubble occurred in 1637 over tulips, when the US federal government was still considered a very remote possibility.
I have to agree with the OP, data is data no matter where it came from. HOWEVER, there is still no point in using the WW2 medical data, for reasons which do not require ethics to explain. Science depends on repeatability at will.
One time data sets can be historically interesting, but if they can't be duplicated today and tomorrow, then they don't strictly add to our understanding of the world. This is the problem with the WWII Nazi experiments. They aren't acceptable and are incompatible with human rights, therefore they can't be duplicated and refined/improved. They are best ignored and left for the historians and philosophers.
I suspect that the big boost in music spending had more to do with the relative affluence of the 60s generation, and their kids in the 80s.