I should preface this by saying that I didn't bother to read anything but the researcher's article. Science journalism is mostly crap; if the original article is available I suggest as a general rule to just head straight for the article.
I wouldn't call it "junk analysis" either, but I'd still put it into the junk science category. Admittedly, IAAP (I am a physicist), but the idea of hanging a long term societal trend on a simple variation in a single--or even a few dozen--variables just smacks of wishful thinking.
It's just that, though- a trend. There's still plenty of error variance to be explained. Reading the paper shows that it doesn't even attempt to explain the greater part of the trend- they explicitly note that there is an as-yet-unaccounted for exponential increase in violent crime over the period they looked at.
And this is not an out-there idea in the first place. Society is made up of people interacting with each other. People act because of how their brains interact with their environments and bodies. People's brains are the way they are because they go through genetically-programmed sequences modulated by environmental influences. Lead is an environmental influence that has been empirically shown to fuck up brains. Fucked-up brain development yields fucked-up brains, which yields fucked up-behavior. So what should happen if we reduce people's exposure to lead?
Let's put it this way- if I told you that I could alter the fine structure constant for some local region of space, you'd probably be shocked. But after you got over that shock, would you be surprised that chemistry operated differently in the altered regions of space? I'd wager probably not. It's to be expected that changes to lower areas in the hierarchy of emergence should produce changes in the higher levels.
An even better test would be going back in history and try to explain the crime rates in Victorian England or other places before the invention of leaded gasoline with this theory. Interestingly enough, there are some historians who claim that the fall of the Roman empire was caused by the increased use of lead pipes in Roman cities.
That would be an interesting test, but it wouldn't necessarily be better, because the data's probably not there at the same level of detail we have for the recent history of the U.S.
So at best, it's shoddy reporting. And as many others here have pointed out, it's probably a journalist trying for one of the oh so popular simple explanations.
I won't argue; I avoid science reporting whenever I can.
I would still side with "lets find some empirical proof that lead damaged children are more criminal than less lead damaged children", obviously taking into account that usually poorer people live closer to the freeways.
So, yes, the story is still out, and even the researchers say that they found intriguing statistical correlations, but no proof whatsoever.
The study was empirical. It wasn't experimental, and it can't be because of ethical considerations. And, like any hypothesis, it can't be proven- just plausibly demonstrated, or disproven.
A really interesting meta study would be if there are scientific fields where the advent of computers have hurt more than they helped. Obviously, my field of nuclear physics has profited enormously--as has this researchers' field of economics, but I can't help feeling that an over reliance on statistical correlations is keeping many from doing real scientific investigation and empirical studies in some fields. In many ways, natural scientists are the lucky ones, because we can experimentally test hypothesis in most cases. Social scientists very rarely have this option.
I don't know how such a study could be carried out objectively. As someone who has worked in a psychology lab, computers didn't seem to be the problem. The problem for us was the tremendous cultural inertia of certain statistical metho
So how would you manipulate the amount of leaded gasoline? The ethical problems with that approach are a major reason why people do correlational research in the first place. Imagine that you suspect that some substance could be a deadly poison. Would you be willing to find out by intentionally administering it?
The correlation != causation argument always forgets that the alternative is a third factor causing BOTH. The idea that correlation is not equal to causation is quite consistent with the idea that correlated variables could share a common cause. I'm not sure anyone that reminds us that correlation is not causation is necessarily forgetting anything.
Reading the article, they already control for abortion, the average crime rate per year, the average crime rate for individual states, and even the effects of people moving from one state to another. The lead level measurements were finer grained than "lead existed before this date, then, everyone stopped using it"- they included state-by-state, year-by-year measurements in their lead data, adjusting for population density (as a surrogate for traffic density).
This was a sophisticated analysis; I wouldn't call it, as some commenters above have, "junk science". It would be surprising for their observed relation to hold, but their interpretation be incorrect. It would be interesting for someone to really come up with an alternative explanation of this paper's observations.
As a side note, I'm pretty sure that by now most lay people, and everyone reading this forum, is aware that correlation does not imply causation. And I'd be willing to guess that the vast majority of scientists have been aware of this elementary statistical fact for some time. It's likely that scientists take many potential influences into account before submitting for publication. So can we please exercise some restraint in the future and actually read the article before denouncing it as "junk science" because, as everyone knows, correlation is not causation? I am emphatically not asking people to take what the researcher says on faith, but if you decry the article without reading it, then your words are essentially noise.
Perhaps your sample of Americans and the state of their health care was affected by the fact that you were getting it from an expensive boat. I suspect that there are other, non-expensive-boat-owning demographics that could have more difficulty paying for their kids' health care.
The reference was to a kind of variable that can be used when programming computers (called a "data type"). Data types for numbers differ in three key ways:
1. How many bits are used to store the number? 2. Is the decimal point free to move (called "floating point", akin to scientific notation), or is it fixed in place (like an integer, whose decimal point is stuck to the right of all the other digits)? 3. Is it possible for instances of this data type to represent both positive and negative numbers, or is this data type restricted to numbers of all the same sign?
A "signed int" refers to a data type that can store integers (positive or negative). Now, to keep track of the sign (ie. whether the variable is positive or negative) takes one bit (ignoring for the moment a slightly more efficient scheme called "two's complement representation"). Spending that one bit to remember what sign the number is reduces the highest absolute value you can store with a signed data type. So an unsigned int might be able to represent numbers from 0 to about 4 billion- but a signed int can represent numbers from negative 2 billion to positive 2 billion.
The sign of the number is stored as the highest bit. So if you add something positive to a very large positive signed int, the carry from the second-to-highest bit will make the highest bit a 1- incorrectly indicating a negative answer. This phenomenon- where two positive signed ints added together give a negative result- is called "overflow".
Re:why check everything
on
Cracking Go
·
· Score: 3, Insightful
By that logic any problem can be called "constant time" as long as you've chosen an actual problem to solve. It's like that, although what I call a problem, you would likely call a class of problems, and what you would call a problem, I would call an instance of a problem. Using your terminology, it's useful to speak of the asymptotic complexity of a class of problems, but not of a single problem.
Just because you've fixed n to some value doesn't change the problem's complexity. If I say that such-and-such a problem can be solved in O(n^2) time, then what I mean by that is that the number of computations required to solve this problem is less than some function of the form an^2 + bn + c for some a, b, and c. A simpler way to say this is that the number of computations required to solve this problem is less than dn^2 (with the right choice of d, this can be greater than an^2 + bn + c for any particular choices of a, b, and c). So, to summarize, O(n^2) time means the number of computations is strictly less than dn^2 for some d (where d does NOT change as problem size changes).
Let's say I am going to bubble-sort a list that is n items long. The most aggressive (lowest) upper bound I can put on this operation is that it will take n(n+1)/2 swaps, so it will take O(n^2) time. Now, let's say that I know that the list is 10 items long. At this point, I can estimate the number of swaps will be at most 10*11/2 = 55.
Now, 55 is O(1), because there is d*1 for some d that is larger than 55 everywhere. Therefore, a bubble-sort run specifically on a list that is 10 items long is constant time.
There is an algorithm to find the median of a list that operates with a surprisingly low time-complexity (Iirc, it's O(n), but I might be wrong) that relies on the fact that a sorting problem of fixed size runs in constant time.
Re:Not brute force, Monte-Carlo !!!
on
Cracking Go
·
· Score: 1
Please forgive my ignorance of this matter, but to me it seems that Monte Carlo methods could have some serious weaknesses in evaluating game positions.
If you are generating random possible endgames from the current move under consideration, then it seems as though certain endgames would end up being more probable to have resulted from actual play and so their contribution to the objective function should be weighted more. But how do we know which endgames are more likely? To get this weighting, it seems as though some intermediate results would have to be obtained.
In the process of generating intermediate board states (and I may be thinking too much in terms of brute force methods) we may miss some crucial moves in our sampling; if a crucial consequence of a move in a shallow ply were missed, then a brute force player would be able to take advantage of that gap in the sampling and beat us.
I haven't read about Monte Carlo methods in the context of games, so feel free to point out what I am missing.
Re:why check everything
on
Cracking Go
·
· Score: 1
Solving a 128-bit cipher is a constant time problem. Solving an n-bit cipher is not.
Big-o notation gives us a bound on the number of computations required to solve a problem (to within a proportionality constant). To solve a 128-bit cipher, you simply enumerate all 2^128 possibilities. That may happen to be a lot of computation, but there still is a constant upper bound on the number of computations (to within a proportionality constant that depends on how hard it is to check any one possibility). If the number of bits were allowed to vary, then we could express the upper bound as a function, like 2^n (where n is the number of bits).
Likewise, solving Go on a 19x19 board is a constant time problem. By having fixed the size of the board, we have placed a limit on the number of possible game trees that board could produce, and while that limit is large, it is a constant. The number of game trees is, however, a function of the board size, so you could say that solving Go on an nxn board is not constant time.
Computerworld reports that KOffice 2.0 will be leaner, faster, and enjoy a cleaner code base than OpenOffice. I'm glad they're setting the bar high for themselves.
I'm not sure I understand how your remark applies. (You may have misread "irreverent" as "irrelevant"?)
Look at Peter Jackson's previous work; I don't doubt that he can do dark. There was a good fit between Peter Jackson's capabilities and the requirements of the work re: LotR. I'm not sure Jackson has proved some general competence in translating the spirit of a book; that hinges on how much we generalize from his success with LotR. Since The Hobbit is a lighter work, I'm not sure he's the right choice to bring The Hobbit to the screen. If he brought the same dark feel to The Hobbit that served him in LotR, that would be specifically contrary to the spirit of the book as I read it.
What makes Jackson the right choice? My experience of The Hobbit was that it was a sillier/ more irreverent book than any of LotR. Its tone was different. I'm not saying he's necessarily the wrong choice, but why would we need the same director?
For the average person game has come to be equated with First Person Shooter. I'm not disagreeing when I say that's the saddest thing I've read all day.
Physical entropy is not the same as information entropy. The use of term "entropy" in information theory is merely inspired by the use of the term "entropy" in physics.
Information entropy is dimensionless- in bits, it is the sum over all events of -P(event)*log_2(P(event)).
Pianos have multiple overtones... but the stiffness of the strings means that those overtones do not form an arithmetic series. See The Physics of Musical Instruments for a discussion (and much much more).
When it comes to this, I am a layman, but my guess would be that if the overtones are not an arithmetic series, they will not be as straightforwardly helpful for pitch recognition as a (more typical) arithmetic series would be.
Peer review and outrageously overpriced journals do not have to go hand in hand. Reviewers do their work for free, so it's not out of the question to have a peer-reviewed journal that happened to be distributed for free to anyone that wanted it.
The worst discrepancy between a just-tempered interval and its corresponding equal-tempered interval is the major third, which at four semitones is 2^(4/12) = 1.259. This is quite perceptibly different from the ideal major third, 1.25. But most of us grew up listening to equal-tempered music, so we're used to it.
I'm not sure what I would call myself. If I had a different set of beliefs about the evolution of the economy, I would probably be libertarian.
Unlike what some others have said here, intelligence doesn't enter into the equation (at least for me). What it boils down to is that I distrust the possibility that power will be concentrated in the hands of a small enough number of entities that they will get away with taking away certain powers of mine that are important to me.
If I thought that telling the government "hands off!" would be enough to prevent that, then I would be a libertarian, but as far as I can tell, a libertarian society simply permits an unacceptable concentration of power in the corporate sector. So there's no good label for me.
Slashdot Mirror
Wanting clarity in communication is orthogonal to liberalism.
I wouldn't call it "junk analysis" either, but I'd still put it into the junk science category. Admittedly, IAAP (I am a physicist), but the idea of hanging a long term societal trend on a simple variation in a single--or even a few dozen--variables just smacks of wishful thinking.
It's just that, though- a trend. There's still plenty of error variance to be explained. Reading the paper shows that it doesn't even attempt to explain the greater part of the trend- they explicitly note that there is an as-yet-unaccounted for exponential increase in violent crime over the period they looked at.
And this is not an out-there idea in the first place. Society is made up of people interacting with each other. People act because of how their brains interact with their environments and bodies. People's brains are the way they are because they go through genetically-programmed sequences modulated by environmental influences. Lead is an environmental influence that has been empirically shown to fuck up brains. Fucked-up brain development yields fucked-up brains, which yields fucked up-behavior. So what should happen if we reduce people's exposure to lead?
Let's put it this way- if I told you that I could alter the fine structure constant for some local region of space, you'd probably be shocked. But after you got over that shock, would you be surprised that chemistry operated differently in the altered regions of space? I'd wager probably not. It's to be expected that changes to lower areas in the hierarchy of emergence should produce changes in the higher levels.
An even better test would be going back in history and try to explain the crime rates in Victorian England or other places before the invention of leaded gasoline with this theory. Interestingly enough, there are some historians who claim that the fall of the Roman empire was caused by the increased use of lead pipes in Roman cities.
That would be an interesting test, but it wouldn't necessarily be better, because the data's probably not there at the same level of detail we have for the recent history of the U.S.
So at best, it's shoddy reporting. And as many others here have pointed out, it's probably a journalist trying for one of the oh so popular simple explanations.
I won't argue; I avoid science reporting whenever I can.
I would still side with "lets find some empirical proof that lead damaged children are more criminal than less lead damaged children", obviously taking into account that usually poorer people live closer to the freeways.
So, yes, the story is still out, and even the researchers say that they found intriguing statistical correlations, but no proof whatsoever.
The study was empirical. It wasn't experimental, and it can't be because of ethical considerations. And, like any hypothesis, it can't be proven- just plausibly demonstrated, or disproven.
A really interesting meta study would be if there are scientific fields where the advent of computers have hurt more than they helped. Obviously, my field of nuclear physics has profited enormously--as has this researchers' field of economics, but I can't help feeling that an over reliance on statistical correlations is keeping many from doing real scientific investigation and empirical studies in some fields. In many ways, natural scientists are the lucky ones, because we can experimentally test hypothesis in most cases. Social scientists very rarely have this option.
I don't know how such a study could be carried out objectively. As someone who has worked in a psychology lab, computers didn't seem to be the problem. The problem for us was the tremendous cultural inertia of certain statistical metho
Reading the article, they already control for abortion, the average crime rate per year, the average crime rate for individual states, and even the effects of people moving from one state to another. The lead level measurements were finer grained than "lead existed before this date, then, everyone stopped using it"- they included state-by-state, year-by-year measurements in their lead data, adjusting for population density (as a surrogate for traffic density).
This was a sophisticated analysis; I wouldn't call it, as some commenters above have, "junk science". It would be surprising for their observed relation to hold, but their interpretation be incorrect. It would be interesting for someone to really come up with an alternative explanation of this paper's observations.
As a side note, I'm pretty sure that by now most lay people, and everyone reading this forum, is aware that correlation does not imply causation. And I'd be willing to guess that the vast majority of scientists have been aware of this elementary statistical fact for some time. It's likely that scientists take many potential influences into account before submitting for publication. So can we please exercise some restraint in the future and actually read the article before denouncing it as "junk science" because, as everyone knows, correlation is not causation? I am emphatically not asking people to take what the researcher says on faith, but if you decry the article without reading it, then your words are essentially noise.
Perhaps your sample of Americans and the state of their health care was affected by the fact that you were getting it from an expensive boat. I suspect that there are other, non-expensive-boat-owning demographics that could have more difficulty paying for their kids' health care.
The reference was to a kind of variable that can be used when programming computers (called a "data type"). Data types for numbers differ in three key ways:
1. How many bits are used to store the number?
2. Is the decimal point free to move (called "floating point", akin to scientific notation), or is it fixed in place (like an integer, whose decimal point is stuck to the right of all the other digits)?
3. Is it possible for instances of this data type to represent both positive and negative numbers, or is this data type restricted to numbers of all the same sign?
A "signed int" refers to a data type that can store integers (positive or negative). Now, to keep track of the sign (ie. whether the variable is positive or negative) takes one bit (ignoring for the moment a slightly more efficient scheme called "two's complement representation"). Spending that one bit to remember what sign the number is reduces the highest absolute value you can store with a signed data type. So an unsigned int might be able to represent numbers from 0 to about 4 billion- but a signed int can represent numbers from negative 2 billion to positive 2 billion.
The sign of the number is stored as the highest bit. So if you add something positive to a very large positive signed int, the carry from the second-to-highest bit will make the highest bit a 1- incorrectly indicating a negative answer. This phenomenon- where two positive signed ints added together give a negative result- is called "overflow".
Let's say I am going to bubble-sort a list that is n items long. The most aggressive (lowest) upper bound I can put on this operation is that it will take n(n+1)/2 swaps, so it will take O(n^2) time. Now, let's say that I know that the list is 10 items long. At this point, I can estimate the number of swaps will be at most 10*11/2 = 55.
Now, 55 is O(1), because there is d*1 for some d that is larger than 55 everywhere. Therefore, a bubble-sort run specifically on a list that is 10 items long is constant time.
There is an algorithm to find the median of a list that operates with a surprisingly low time-complexity (Iirc, it's O(n), but I might be wrong) that relies on the fact that a sorting problem of fixed size runs in constant time.
Please forgive my ignorance of this matter, but to me it seems that Monte Carlo methods could have some serious weaknesses in evaluating game positions.
If you are generating random possible endgames from the current move under consideration, then it seems as though certain endgames would end up being more probable to have resulted from actual play and so their contribution to the objective function should be weighted more. But how do we know which endgames are more likely? To get this weighting, it seems as though some intermediate results would have to be obtained.
In the process of generating intermediate board states (and I may be thinking too much in terms of brute force methods) we may miss some crucial moves in our sampling; if a crucial consequence of a move in a shallow ply were missed, then a brute force player would be able to take advantage of that gap in the sampling and beat us.
I haven't read about Monte Carlo methods in the context of games, so feel free to point out what I am missing.
Solving a 128-bit cipher is a constant time problem. Solving an n-bit cipher is not.
Big-o notation gives us a bound on the number of computations required to solve a problem (to within a proportionality constant). To solve a 128-bit cipher, you simply enumerate all 2^128 possibilities. That may happen to be a lot of computation, but there still is a constant upper bound on the number of computations (to within a proportionality constant that depends on how hard it is to check any one possibility). If the number of bits were allowed to vary, then we could express the upper bound as a function, like 2^n (where n is the number of bits).
Likewise, solving Go on a 19x19 board is a constant time problem. By having fixed the size of the board, we have placed a limit on the number of possible game trees that board could produce, and while that limit is large, it is a constant. The number of game trees is, however, a function of the board size, so you could say that solving Go on an nxn board is not constant time.
Well, how else are the whalers going to get there?!?
Iirc, Cerebrospinal fluid does more cooling for the brain than blood does.
I'm not sure I understand how your remark applies. (You may have misread "irreverent" as "irrelevant"?)
Look at Peter Jackson's previous work; I don't doubt that he can do dark. There was a good fit between Peter Jackson's capabilities and the requirements of the work re: LotR. I'm not sure Jackson has proved some general competence in translating the spirit of a book; that hinges on how much we generalize from his success with LotR. Since The Hobbit is a lighter work, I'm not sure he's the right choice to bring The Hobbit to the screen. If he brought the same dark feel to The Hobbit that served him in LotR, that would be specifically contrary to the spirit of the book as I read it.
What makes Jackson the right choice? My experience of The Hobbit was that it was a sillier/ more irreverent book than any of LotR. Its tone was different. I'm not saying he's necessarily the wrong choice, but why would we need the same director?
Oh yeah?!? Well my employer has been working to modernize the health record since 1979!
But a larger talent pool could help, no?
http://xkcd.com/294/
The people here seem to think they can judge the quality of a study based on what they read of it in popular news coverage.
Physical entropy is not the same as information entropy. The use of term "entropy" in information theory is merely inspired by the use of the term "entropy" in physics.
Information entropy is dimensionless- in bits, it is the sum over all events of -P(event)*log_2(P(event)).
Pianos have multiple overtones... but the stiffness of the strings means that those overtones do not form an arithmetic series. See The Physics of Musical Instruments for a discussion (and much much more).
When it comes to this, I am a layman, but my guess would be that if the overtones are not an arithmetic series, they will not be as straightforwardly helpful for pitch recognition as a (more typical) arithmetic series would be.
Peer review and outrageously overpriced journals do not have to go hand in hand. Reviewers do their work for free, so it's not out of the question to have a peer-reviewed journal that happened to be distributed for free to anyone that wanted it.
The worst discrepancy between a just-tempered interval and its corresponding equal-tempered interval is the major third, which at four semitones is 2^(4/12) = 1.259. This is quite perceptibly different from the ideal major third, 1.25. But most of us grew up listening to equal-tempered music, so we're used to it.
I'm not sure what I would call myself. If I had a different set of beliefs about the evolution of the economy, I would probably be libertarian.
Unlike what some others have said here, intelligence doesn't enter into the equation (at least for me). What it boils down to is that I distrust the possibility that power will be concentrated in the hands of a small enough number of entities that they will get away with taking away certain powers of mine that are important to me.
If I thought that telling the government "hands off!" would be enough to prevent that, then I would be a libertarian, but as far as I can tell, a libertarian society simply permits an unacceptable concentration of power in the corporate sector. So there's no good label for me.