That's a different problem. Like this one, it's not a problem with p-values, it's a problem with people who don't know what a p-value is. The examples in the comic are NOT p-values for the experiment that was done. Properly calculated p-values do not have this problem because they are corrected for multiple comparisons.
Hubbard said the best way to get rich is to found a religion. Then he did it. Crazy like a fox?
This guy kinda looks like a successful confidence man who put one over on the world and laughed his way to his fleet of yachts, laden with wide-eyed young worshippers. Oh right, that's exactly what he did.
The King James bible was translated at a king's (god's appointed representative) order, by translators who were divinely inspired. Or so they said. Believing in it is no more irrational than believing in the actual original accounts, verbal or written, or the Hebrew copies, or the Greek copies, or the Book of Mormon, or Hubbard's science fiction. Okay, maybe slightly less irrational than believing in that last one, because Hubbard declared in advance he was full of shit rather than claiming to have a direct pipeline to a supreme being. Or maybe not.
I was kidding, but more seriously, I don't think the satellite connections are going to be much of a problem. Just like the old air phones, most of the flight won't be able to afford it. If they ever start using ground stations to provide affordable links, then hell will break loose.
The situation isn't quite analogous. There are laws that allow the theatre to throw the loudmouth out. For airplanes, there are laws that say they can't.
I did once. When they were new. The conversation was (paraphrased) "Dude! Guess where I'm calling from? Yeah, a plane! What? Shit, this costs HOW much?"
Think and read carefully. I said that using a spherical approximation, which gives a cube rule, and is the basis for the OPs attack on BMI, was not a good approximation of a human being. I used a cylinder, which has a linear relationship to height, as a counter example. This is also not a particularly good approximation. I didn't say it was. It IS better than a sphere.
The square function that BMI follows is a reasonable approximation across the normal height range of human beings. No, it doesn't work particularly well near the edges of that range. No approximation does. BMI is supposed to be used for aggregate statistics, where it works reasonably well. It also does a fair job when applied to individuals, provided those individuals aren't too far away from the mean. Yes, you can pick ridiculous examples that are far from the mean.
So we've established that people aren't spherical (a cube law) and they're not fixed radius cylinders (a linear law). What exponent would you suggest? Something in the middle maybe? Like 2? Do you have any actual evidence for using anything else?
Why are you getting so worked up about a simple approximation?
The volume of a cylinder is linear with height and square with radius. The average person's radius doesn't (or isn't supposed to) scale proportionally with their height, so weight shouldn't follow a cubic formula. Yes, BMI is a very simplified rule of thumb that unfortunately gets treated like a highly prognostic measurement, but on average it works reasonably well. There are better metrics, although they're slightly to a lot more complicated.
In a population the BMI does predict things. Such as heart disease. In individuals it often works poorly. Hip to waist ratio works a bit better in populations and quite a bit better in individuals. A CT scan combined with a displacement body volume measure is about the best thing we've got for individuals, and would also work wonderfully in a population except for the expense and radiation exposure.
BMI isn't a great metric to use here, but it's not hopelessly wrong. Waist to hip ratio would give better data except that more people are probably familiar with their height and weight than waist and hip measurements. Realistically, since it's an opt in program, the BMI is really only protecting them from claims that they're nagging anorexics to lose weight.
Practicing research scientist, at a real university, actually. You make your living with science hey? That could mean anything. From your tone I formulated a few hypotheses.
When two models (some people call these "theories" or "hypotheses") fit the data, but one fits better, we call that evidence in favour of that model (which you agree with). The simplest model that fits the data well (i.e. the simplest model that fits the most data) is considered the best description of reality. It's probably wrong, at least in some details, but it's the best at the moment. That's as close as anyone can get to "describing reality."
Did you do the experiment in high school where you measure a ball rolling down an inclined plane and plot the position over time? Then you fit a curve to the points and come up with d = vit+1/2at^2? That experiment is evidence that objects undergoing accelerated motion move in a way prescribed by that equation. Yes, it's not fully correct in this particular case (which you can detect in high school if you do it carefully enough). The angular velocity of the rolling ball soaks up enough energy to be noticeable. Also friction, of course.
What these people (other practicing research scientists at real universities, by the way) have done is similar. They figured out what pattern of mistakes would be expected from classical thermal annealing and quantum annealing (using models called thermodynamics and quantum mechanics), and compared these predictions to what the machine actually outputs. One fits better, providing evidence that the machine is more likely using that process.
Just like all the other electric cars on the market, the Tesla is a luxury commuter car for city use. Since the story is about Tesla making claims that it's a viable drop in replacement for a gasoline car, specifically that it can do cross country road trips, the criticism is more than fair.
For those of you in the know, the question isn't speed but the rate of growth of the speed: is the ratio of speed-up growing polynomially in the input, or exponentially?
It isn't even really that. Quantum annealing is faster than thermal annealing in some situations (tall thin boundaries between local minima) and slower in others (every other solution topography). Whether or not D-Wave's computer is quantum or not seems to be mostly a question of whether the marketing department can continue using the magic word.
The D-Wave computer is supposed to perform quantum annealing. Annealing, quantum or otherwise, is essentially an optimization problem (you're looking for the global minimum). Both the quantum and the classical annealing operations will find local (but not global) minima some of the time, but the pattern of these non-optimal solutions is different between the two. So you take the output from the D-Wave and try to match it to simulations of quantum and classical annealing. According to the article, the D-Wave output more closely matched what you'd expect from simulated quantum annealing but these researchers have now come up with a simulation of classical annealing that more closely matches.
If I understand it correctly, this thing is designed to simulate annealing. Annealing is he process the happens when you slowly cool a metal alloy: atoms are bouncing around with lots of thermal energy at first, then less and less. As you cool them, they're more likely to get stuck in low energy configurations rather than high. You can do most non-convex (and convex I suppose, but there are better ways to solve those) optimization problems using simulated annealing.
Simulated annealing (or quasi-simulated for a quantum computer I guess) should be something quantum computers are really good at doing quickly. Where particles (simulated or otherwise) in a classical system have to have enough thermal energy to jump out of local minima, quantum particles can tunnel out, so quantum annealing can work better for particular types of problems. But since a quantum computer can do the annealing physically, in massive parallel, it's potentially much faster than serial simulation.
Despite all the hype about factoring, fast non-convex optimization might be QC's biggest killer app. D-Wave's computer DOES solve annealing problems. The question is, does it solve them like a classical computer does, with particles jumping out of local minima, or like a quantum system, with tunnelling? If nothing else, maybe the D-Wave controversy has spurred more fast non-convex optimization research.
The same thing occurred to me. The thing is, DWave's chip definitely doesn't have a bunch of transistors on it executing a conventional program. It's more like a little physical simulation on a chip. The question is whether that particular setup behaves classically or not.
The people who lost those ships insist, vehemently, that they don't enjoy building them. I kind of doubt that's true, or at least doubt it's as true as they make it out to be.
I agree the wealth calculation is flawed, but it's at least as valid as most other commonly used wealth measurements. "$300,000 Destroyed in Internet Spaceship Battle" is no worse than "Amazon Loses $10 Million a Minute Due to DDOS" or "$100 Billion A Year Losses from Worker Sick Days." All of those are just big numbers we enjoy seeing in print because we're monkeys who like shiny things.
Slashdot Mirror
That's a different problem. Like this one, it's not a problem with p-values, it's a problem with people who don't know what a p-value is. The examples in the comic are NOT p-values for the experiment that was done. Properly calculated p-values do not have this problem because they are corrected for multiple comparisons.
Hubbard said the best way to get rich is to found a religion. Then he did it. Crazy like a fox?
This guy kinda looks like a successful confidence man who put one over on the world and laughed his way to his fleet of yachts, laden with wide-eyed young worshippers. Oh right, that's exactly what he did.
The King James bible was translated at a king's (god's appointed representative) order, by translators who were divinely inspired. Or so they said. Believing in it is no more irrational than believing in the actual original accounts, verbal or written, or the Hebrew copies, or the Greek copies, or the Book of Mormon, or Hubbard's science fiction. Okay, maybe slightly less irrational than believing in that last one, because Hubbard declared in advance he was full of shit rather than claiming to have a direct pipeline to a supreme being. Or maybe not.
I was kidding, but more seriously, I don't think the satellite connections are going to be much of a problem. Just like the old air phones, most of the flight won't be able to afford it. If they ever start using ground stations to provide affordable links, then hell will break loose.
PUS? I don't think they exist.
The situation isn't quite analogous. There are laws that allow the theatre to throw the loudmouth out. For airplanes, there are laws that say they can't.
I don't think a cell phone would work very well on the over-seas portion of a flight anyway. Poor service out there.
I did once. When they were new. The conversation was (paraphrased) "Dude! Guess where I'm calling from? Yeah, a plane! What? Shit, this costs HOW much?"
If you were a phone company why would you want to lobby against people using your service, and probably paying ridiculous roaming fees while doing it?
Now, if you were an airline you might lobby for this so you didn't have to deal with brawls on your planes.
Think and read carefully. I said that using a spherical approximation, which gives a cube rule, and is the basis for the OPs attack on BMI, was not a good approximation of a human being. I used a cylinder, which has a linear relationship to height, as a counter example. This is also not a particularly good approximation. I didn't say it was. It IS better than a sphere.
The square function that BMI follows is a reasonable approximation across the normal height range of human beings. No, it doesn't work particularly well near the edges of that range. No approximation does. BMI is supposed to be used for aggregate statistics, where it works reasonably well. It also does a fair job when applied to individuals, provided those individuals aren't too far away from the mean. Yes, you can pick ridiculous examples that are far from the mean.
So we've established that people aren't spherical (a cube law) and they're not fixed radius cylinders (a linear law). What exponent would you suggest? Something in the middle maybe? Like 2? Do you have any actual evidence for using anything else?
Why are you getting so worked up about a simple approximation?
You're aware that people aren't spherical, right?
(Okay, I guess some people approach spherical.)
The volume of a cylinder is linear with height and square with radius. The average person's radius doesn't (or isn't supposed to) scale proportionally with their height, so weight shouldn't follow a cubic formula. Yes, BMI is a very simplified rule of thumb that unfortunately gets treated like a highly prognostic measurement, but on average it works reasonably well. There are better metrics, although they're slightly to a lot more complicated.
In a population the BMI does predict things. Such as heart disease. In individuals it often works poorly. Hip to waist ratio works a bit better in populations and quite a bit better in individuals. A CT scan combined with a displacement body volume measure is about the best thing we've got for individuals, and would also work wonderfully in a population except for the expense and radiation exposure.
BMI isn't a great metric to use here, but it's not hopelessly wrong. Waist to hip ratio would give better data except that more people are probably familiar with their height and weight than waist and hip measurements. Realistically, since it's an opt in program, the BMI is really only protecting them from claims that they're nagging anorexics to lose weight.
Practicing research scientist, at a real university, actually. You make your living with science hey? That could mean anything. From your tone I formulated a few hypotheses.
When two models (some people call these "theories" or "hypotheses") fit the data, but one fits better, we call that evidence in favour of that model (which you agree with). The simplest model that fits the data well (i.e. the simplest model that fits the most data) is considered the best description of reality. It's probably wrong, at least in some details, but it's the best at the moment. That's as close as anyone can get to "describing reality."
Did you do the experiment in high school where you measure a ball rolling down an inclined plane and plot the position over time? Then you fit a curve to the points and come up with d = vit+1/2at^2? That experiment is evidence that objects undergoing accelerated motion move in a way prescribed by that equation. Yes, it's not fully correct in this particular case (which you can detect in high school if you do it carefully enough). The angular velocity of the rolling ball soaks up enough energy to be noticeable. Also friction, of course.
What these people (other practicing research scientists at real universities, by the way) have done is similar. They figured out what pattern of mistakes would be expected from classical thermal annealing and quantum annealing (using models called thermodynamics and quantum mechanics), and compared these predictions to what the machine actually outputs. One fits better, providing evidence that the machine is more likely using that process.
Just like all the other electric cars on the market, the Tesla is a luxury commuter car for city use. Since the story is about Tesla making claims that it's a viable drop in replacement for a gasoline car, specifically that it can do cross country road trips, the criticism is more than fair.
The web site wouldn't let me match my typical winter driving conditions, but the closest I could get put the range at 171 miles.
Some people don't really want to check into a hotel for the night when they pull into a fuel station.
It isn't even really that. Quantum annealing is faster than thermal annealing in some situations (tall thin boundaries between local minima) and slower in others (every other solution topography). Whether or not D-Wave's computer is quantum or not seems to be mostly a question of whether the marketing department can continue using the magic word.
The D-Wave computer is supposed to perform quantum annealing. Annealing, quantum or otherwise, is essentially an optimization problem (you're looking for the global minimum). Both the quantum and the classical annealing operations will find local (but not global) minima some of the time, but the pattern of these non-optimal solutions is different between the two. So you take the output from the D-Wave and try to match it to simulations of quantum and classical annealing. According to the article, the D-Wave output more closely matched what you'd expect from simulated quantum annealing but these researchers have now come up with a simulation of classical annealing that more closely matches.
Except for all of scientific knowledge, of course.
It's pretty clear that you, just like most of the other posters, don't understand what D-Wave's machine does or how it's being tested.
Ah, if only it were the 90s again.
If I understand it correctly, this thing is designed to simulate annealing. Annealing is he process the happens when you slowly cool a metal alloy: atoms are bouncing around with lots of thermal energy at first, then less and less. As you cool them, they're more likely to get stuck in low energy configurations rather than high. You can do most non-convex (and convex I suppose, but there are better ways to solve those) optimization problems using simulated annealing.
Simulated annealing (or quasi-simulated for a quantum computer I guess) should be something quantum computers are really good at doing quickly. Where particles (simulated or otherwise) in a classical system have to have enough thermal energy to jump out of local minima, quantum particles can tunnel out, so quantum annealing can work better for particular types of problems. But since a quantum computer can do the annealing physically, in massive parallel, it's potentially much faster than serial simulation.
Despite all the hype about factoring, fast non-convex optimization might be QC's biggest killer app. D-Wave's computer DOES solve annealing problems. The question is, does it solve them like a classical computer does, with particles jumping out of local minima, or like a quantum system, with tunnelling? If nothing else, maybe the D-Wave controversy has spurred more fast non-convex optimization research.
The same thing occurred to me. The thing is, DWave's chip definitely doesn't have a bunch of transistors on it executing a conventional program. It's more like a little physical simulation on a chip. The question is whether that particular setup behaves classically or not.
With a match.
Stallman whining isn't a problem for me. What makes you think it is?
Or do you always conflate disagreement with a problem?
The people who lost those ships insist, vehemently, that they don't enjoy building them. I kind of doubt that's true, or at least doubt it's as true as they make it out to be.
I agree the wealth calculation is flawed, but it's at least as valid as most other commonly used wealth measurements. "$300,000 Destroyed in Internet Spaceship Battle" is no worse than "Amazon Loses $10 Million a Minute Due to DDOS" or "$100 Billion A Year Losses from Worker Sick Days." All of those are just big numbers we enjoy seeing in print because we're monkeys who like shiny things.