|
|
|
Thread Tools | Display Modes |
#11
|
|||
|
|||
Galaxies without dark matter halos?
greywolf42 wrote:
Ralph Hartley wrote: Huh??? The (Bayesian) probability of the big bang is very close to 1. Depending on how you rate the evidence for inflation the relative probabilities with and without inflation will vary, but still sum to near 1 (they are mutually exclusive by construction). The Bayesian probability is whatever you set the prior. Modified by the evidence you have seen so far. If you pick a good prior, the evidence will eventually win over your prejudices. Since you have defined your 'prior' as including only variants of the Big Bang. (You didn't mention any non-Big Bang theories that you addressed -- thus setting the prior for all big-bang variants to 1.0.) No. I'm saying that the *posteriori* probabilities are that way. I haven't seen any Non-BB theories that aren't contradicted by evidence. None have enough a-posteriori probability to be worth bothering with. Well, we've seen galaxies that have no hint of dark matter. This would require yet another ad hoc revision to the model -- allowing one to randomly distribute 'dark matter' wherever one finds a discrepant observation -- and withhold dark matter where everything looks fine. Non Big Bang theories (if there were any worth mentioning) would have to do the same thing, so it's a wash. If you don't believe in dark matter, you still need to explain why some galaxies behave *as* *if* they had dark matter and some do not. I don't see how eliminating the Big Bang helps you do that. I'm talking about non-Bayesian statistics. I don't see what "the non-Bayesian probability of the Big Bang" even *means*! The Big Bang ether happened or it didn't. I can express my knowledge as a probability (basically what odds I would consider a fair bet), that's what Bayesian probability means. What do *you* mean by "the probability of the big bang"? Hence -- since it is 'open' -- we have no way to set the probability. Therefore, since one cannot define a real (non-Bayesian) probability for the (qualitatively) 'most likely' BB theory -- we cannot define a real (non-Bayesian) probability of "near certainty" (arbitrarily close to 1.0 in a NON-bayesian sense) to the existence of the Big Bang explosion. NON-Bayesian probability doesn't apply *at* *all* to unique events, so why would I want to do such a thing? Nor can evidence for or against different theories for how the earth formed be taken as evidence for or against the "round earth" theory. It depends on which 'round Earth' theory you are championing. There are many. (Expanding Earth, plate tectonics, immoble Earth, shrinking crust....) Some 'Earth formation' theories even include a non-round Earth. I generally believe in the "earth is pretty much round" theory, which includes all of those. I will bet you a dollar that, when all is said and done, it will turn out that the earth is more or less round. I will give you 100:1 odds (i.e. my odds for a round earth are more than 0.99). I was not born with this belief, but having seen lots of evidence for it, hold it fairly strongly, and would only be convinced otherwise by sailing over the edge. Show me the edge, and I'll give you $100, but forgive me for asking to hold your $1 in the meantime. Similarly, the Big Bang describes the rough shape of the universe. Neither arguments about its exact shape (e.g. the value of Omega), nor about its origin (inflation or not) make any difference. Of course they make a difference to the overall probability! How? Why? And if you reject Bayesian probability what do you even *mean*. The problem is that you are starting with a Bayesian prior that the probability of the Big Bang is arbitrarily close to 1.0. No. I had to be convinced. I was. Ralph Hartley |
#12
|
|||
|
|||
Galaxies without dark matter halos?
Ralph Hartley wrote in message
... greywolf42 wrote: Ralph Hartley wrote: Huh??? The (Bayesian) probability of the big bang is very close to 1. Depending on how you rate the evidence for inflation the relative probabilities with and without inflation will vary, but still sum to near 1 (they are mutually exclusive by construction). The Bayesian probability is whatever you set the prior. Modified by the evidence you have seen so far. If you pick a good prior, the evidence will eventually win over your prejudices. The key word being "eventually." Here there is only one "data" point -- the prior. Which is why Bayesian statistics are invalid here. Since you have defined your 'prior' as including only variants of the Big Bang. (You didn't mention any non-Big Bang theories that you addressed -- thus setting the prior for all big-bang variants to 1.0.) No. I'm saying that the *posteriori* probabilities are that way. I haven't seen any Non-BB theories that aren't contradicted by evidence. None have enough a-posteriori probability to be worth bothering with. Regardless of the probabilities that you set on the theories that *you* have seen, they do exist. By simply dismissing them without thought, you assume the BB is correct. A priori, the 'correct' theory of the cosmos may not have yet been found. Yet it's probability must be allowed for. If you want, you can 'rate' the probability of one known theory against another known theory. But you can't claim the BB is a 'near certainty' -- unless you assume it by attempting to apply a Bayesian approach. Well, we've seen galaxies that have no hint of dark matter. This would require yet another ad hoc revision to the model -- allowing one to randomly distribute 'dark matter' wherever one finds a discrepant observation -- and withhold dark matter where everything looks fine. Non Big Bang theories (if there were any worth mentioning) would have to do the same thing, so it's a wash. Non-big bang theories do not require 'dark matter.' This ever-shifting dark matter is a requirement only for big-bang theories. If you don't believe in dark matter, you still need to explain why some galaxies behave *as* *if* they had dark matter and some do not. I don't see how eliminating the Big Bang helps you do that. The primary difference for the BB is the difference between non-baryonic 'dark matter' (required by the BB to keep the light elements 'cooked' right) and just as-yet-unseen-because-it's-not-highly-luminous normal matter (sometimes confusingly called dark matter) that would be used by non-BB theories. I consider (not 'believe') dark matter to be a phantom. It is an ad hoc speculation made simply to 'rescue' a popular theory from observation. I'm talking about non-Bayesian statistics. I don't see what "the non-Bayesian probability of the Big Bang" even *means*! The Big Bang ether happened or it didn't. I can express my knowledge as a probability (basically what odds I would consider a fair bet), that's what Bayesian probability means. What do *you* mean by "the probability of the big bang"? Care to play some 3-card monte? I'll give you the Bayesian odds. You *think* you know the odds (at worst 1/3, better if your eye is quick) -- because your theory doesn't (yet) include reality. So you set up your Bayesian prior, and happily wait for the payoff. Somehow, you ALWAYS lose. Because you didn't have the correct theory ready to hand. You had *a* theory. Not *all* theories. That is the difference between reality and Bayesian priors. Hence -- since it is 'open' -- we have no way to set the probability. Therefore, since one cannot define a real (non-Bayesian) probability for the (qualitatively) 'most likely' BB theory -- we cannot define a real (non-Bayesian) probability of "near certainty" (arbitrarily close to 1.0 in a NON-bayesian sense) to the existence of the Big Bang explosion. NON-Bayesian probability doesn't apply *at* *all* to unique events, so why would I want to do such a thing? We are applying probability to cosmological *theories* -- not to a theoretical event contained within only SOME of those theories. If I use the other theories, then the Bayesian probability of the BB event is zero. (The BB event is not unique if it never occurred.) Again, we can compare any two theories by relative probability (this does not require Bayesian probability). And we can arbitrarily select the 'most probable' theory and call it a '1.0' probability. But these probabilities are all *relative* and will always be relative. One cannot invoke a Bayesian prior (which is an arbitrary choice) and then use this arbitrary choice to claim an event within the theories is a 'near certainty.' Nor can evidence for or against different theories for how the earth formed be taken as evidence for or against the "round earth" theory. It depends on which 'round Earth' theory you are championing. There are many. (Expanding Earth, plate tectonics, immoble Earth, shrinking crust....) Some 'Earth formation' theories even include a non-round Earth. I generally believe in the "earth is pretty much round" theory, which includes all of those. It obviously doesn't include the 'non-round' Earth. 'Pretty much round' is not a theory. I see no cause-and-effect, which is required for a scientific theory. It's not even quantified. I will bet you a dollar that, when all is said and done, it will turn out that the earth is more or less round. I will give you 100:1 odds (i.e. my odds for a round earth are more than 0.99). Do I get to define 'more or less round?' Without quantification on your part, you've set up a sucker bet. (I really would like to deal for some three card monte.) I was not born with this belief, but having seen lots of evidence for it, holdit fairly strongly, and would only be convinced otherwise by sailing over the edge. Show me the edge, and I'll give you $100, but forgive me for asking to hold your $1 in the meantime. There are 'non-round' solutions without edges. Similarly, the Big Bang describes the rough shape of the universe. Neither arguments about its exact shape (e.g. the value of Omega), nor about its origin (inflation or not) make any difference. Of course they make a difference to the overall probability! How? Why? And if you reject Bayesian probability what do you even *mean*. If the BB describes the shape of the universe, then observations of the shape of the universe (the value of omega) are the fundamental means for checking the theory. Therefore, they make a difference. The problem is that you are starting with a Bayesian prior that the probability of the Big Bang is arbitrarily close to 1.0. No. I had to be convinced. I was. As described above, to be 'convinced' you used a Bayesian prior of 1.0. You are not allowing for your (or our) ignorance. greywolf42 ubi dubium ibi libertas |
#13
|
|||
|
|||
Galaxies without dark matter halos?
Ralph Hartley wrote in message
... greywolf42 wrote: Ralph Hartley wrote: Huh??? The (Bayesian) probability of the big bang is very close to 1. Depending on how you rate the evidence for inflation the relative probabilities with and without inflation will vary, but still sum to near 1 (they are mutually exclusive by construction). The Bayesian probability is whatever you set the prior. Modified by the evidence you have seen so far. If you pick a good prior, the evidence will eventually win over your prejudices. The key word being "eventually." Here there is only one "data" point -- the prior. Which is why Bayesian statistics are invalid here. Since you have defined your 'prior' as including only variants of the Big Bang. (You didn't mention any non-Big Bang theories that you addressed -- thus setting the prior for all big-bang variants to 1.0.) No. I'm saying that the *posteriori* probabilities are that way. I haven't seen any Non-BB theories that aren't contradicted by evidence. None have enough a-posteriori probability to be worth bothering with. Regardless of the probabilities that you set on the theories that *you* have seen, they do exist. By simply dismissing them without thought, you assume the BB is correct. A priori, the 'correct' theory of the cosmos may not have yet been found. Yet it's probability must be allowed for. If you want, you can 'rate' the probability of one known theory against another known theory. But you can't claim the BB is a 'near certainty' -- unless you assume it by attempting to apply a Bayesian approach. Well, we've seen galaxies that have no hint of dark matter. This would require yet another ad hoc revision to the model -- allowing one to randomly distribute 'dark matter' wherever one finds a discrepant observation -- and withhold dark matter where everything looks fine. Non Big Bang theories (if there were any worth mentioning) would have to do the same thing, so it's a wash. Non-big bang theories do not require 'dark matter.' This ever-shifting dark matter is a requirement only for big-bang theories. If you don't believe in dark matter, you still need to explain why some galaxies behave *as* *if* they had dark matter and some do not. I don't see how eliminating the Big Bang helps you do that. The primary difference for the BB is the difference between non-baryonic 'dark matter' (required by the BB to keep the light elements 'cooked' right) and just as-yet-unseen-because-it's-not-highly-luminous normal matter (sometimes confusingly called dark matter) that would be used by non-BB theories. I consider (not 'believe') dark matter to be a phantom. It is an ad hoc speculation made simply to 'rescue' a popular theory from observation. I'm talking about non-Bayesian statistics. I don't see what "the non-Bayesian probability of the Big Bang" even *means*! The Big Bang ether happened or it didn't. I can express my knowledge as a probability (basically what odds I would consider a fair bet), that's what Bayesian probability means. What do *you* mean by "the probability of the big bang"? Care to play some 3-card monte? I'll give you the Bayesian odds. You *think* you know the odds (at worst 1/3, better if your eye is quick) -- because your theory doesn't (yet) include reality. So you set up your Bayesian prior, and happily wait for the payoff. Somehow, you ALWAYS lose. Because you didn't have the correct theory ready to hand. You had *a* theory. Not *all* theories. That is the difference between reality and Bayesian priors. Hence -- since it is 'open' -- we have no way to set the probability. Therefore, since one cannot define a real (non-Bayesian) probability for the (qualitatively) 'most likely' BB theory -- we cannot define a real (non-Bayesian) probability of "near certainty" (arbitrarily close to 1.0 in a NON-bayesian sense) to the existence of the Big Bang explosion. NON-Bayesian probability doesn't apply *at* *all* to unique events, so why would I want to do such a thing? We are applying probability to cosmological *theories* -- not to a theoretical event contained within only SOME of those theories. If I use the other theories, then the Bayesian probability of the BB event is zero. (The BB event is not unique if it never occurred.) Again, we can compare any two theories by relative probability (this does not require Bayesian probability). And we can arbitrarily select the 'most probable' theory and call it a '1.0' probability. But these probabilities are all *relative* and will always be relative. One cannot invoke a Bayesian prior (which is an arbitrary choice) and then use this arbitrary choice to claim an event within the theories is a 'near certainty.' Nor can evidence for or against different theories for how the earth formed be taken as evidence for or against the "round earth" theory. It depends on which 'round Earth' theory you are championing. There are many. (Expanding Earth, plate tectonics, immoble Earth, shrinking crust....) Some 'Earth formation' theories even include a non-round Earth. I generally believe in the "earth is pretty much round" theory, which includes all of those. It obviously doesn't include the 'non-round' Earth. 'Pretty much round' is not a theory. I see no cause-and-effect, which is required for a scientific theory. It's not even quantified. I will bet you a dollar that, when all is said and done, it will turn out that the earth is more or less round. I will give you 100:1 odds (i.e. my odds for a round earth are more than 0.99). Do I get to define 'more or less round?' Without quantification on your part, you've set up a sucker bet. (I really would like to deal for some three card monte.) I was not born with this belief, but having seen lots of evidence for it, holdit fairly strongly, and would only be convinced otherwise by sailing over the edge. Show me the edge, and I'll give you $100, but forgive me for asking to hold your $1 in the meantime. There are 'non-round' solutions without edges. Similarly, the Big Bang describes the rough shape of the universe. Neither arguments about its exact shape (e.g. the value of Omega), nor about its origin (inflation or not) make any difference. Of course they make a difference to the overall probability! How? Why? And if you reject Bayesian probability what do you even *mean*. If the BB describes the shape of the universe, then observations of the shape of the universe (the value of omega) are the fundamental means for checking the theory. Therefore, they make a difference. The problem is that you are starting with a Bayesian prior that the probability of the Big Bang is arbitrarily close to 1.0. No. I had to be convinced. I was. As described above, to be 'convinced' you used a Bayesian prior of 1.0. You are not allowing for your (or our) ignorance. greywolf42 ubi dubium ibi libertas |
#14
|
|||
|
|||
Galaxies without dark matter halos?
greywolf42 wrote:
Ralph Hartley wrote: greywolf42 wrote: .... The Bayesian probability is whatever you set the prior. Modified by the evidence you have seen so far. If you pick a good prior, the evidence will eventually win over your prejudices. The key word being "eventually." Here there is only one "data" point -- the prior. Which is why Bayesian statistics are invalid here. Huh? Has all astronomy in the past century not provided any data at all? Before I heard of (I wasn't born when it was discovered) the Hubble law I would have set the probability of the BB (if I had thought of it at all) to maybe P_prior(BB)~=0.05. That's my prior. Don't say it is 1. It is not. Until I read about the microwave background, I preferred the Steady State universe. Say P_hubble(BB)~=0.25. Even after that I wasn't completely convinced (their frequency coverage wasn't that good) P_cmb(BB)~=0.7. It only moved into "practically certain" territory when I saw the COBE data. Since then lots of data has come in. Light element abundances (I think this was known before COBE but I heard of it later), gravitational lensing, apparent evolution (distant galaxies etc. look different), CMB anisotropy, etc.. All of which have enough parameters that they don't' *have* to match *any* Big Bang variant, but they *do*. I stop keeping track at 99%. Do you actually *know* of a non-BB theory that predicts the black body CMB? What fluctuations does it predict? What element abundances? What red shifts? .... The primary difference for the BB is the difference between non-baryonic 'dark matter' (required by the BB to keep the light elements 'cooked' right) and just as-yet-unseen-because-it's-not-highly-luminous normal matter (sometimes confusingly called dark matter) that would be used by non-BB theories. Baryonic or not, you have to explain why some galaxies have it and some (assuming, for the sake of argument, that the report was correct) do not. .... NON-Bayesian probability doesn't apply *at* *all* to unique events, so why would I want to do such a thing? We are applying probability to cosmological *theories* -- not to a theoretical event contained within only SOME of those theories. If I use the other theories, then the Bayesian probability of the BB event is zero. It doesn't matter what theory you *use*, only which is *right*. There is no meaningful way to apply non-Bayesian probability to theories, since only one is right, and it's always the same one. If someone says "probability of the Big Bang" they are either talking nonsense, or talking about Bayesian probability. When I assumed you were talking about Bayesian probability I was giving you the benefit of the doubt, maybe that was wrong:-). Again, we can compare any two theories by relative probability (this does not require Bayesian probability). It does, even if you don't admit it. And we can arbitrarily select the 'most probable' theory and call it a '1.0' probability. We could, but that would be wrong, and I never did that. The probability of the best theory might not be even close to 1. Especially if there is a continuum of possible theories (there usually is). A set of theories (e.g. the set of BB theories) can have a combined probability near 1 even if every member of the set has infinitesimal probability. A good Bayesian never sets anything to 1. I said it was *close* to 1. ... One cannot invoke a Bayesian prior (which is an arbitrary choice) and then use this arbitrary choice to claim an event within the theories is a 'near certainty.' Of course I don't use my prior of 0.05 to claim that it is a "near certainty". I use my posterior of 0.99+. .... Do I get to define 'more or less round?' Within reasonable limits. If you can't guess approximately what I would consider reasonable, there is not much point in arguing. Tell the truth, you *do* know what I mean when I say "the earth is round", don't you? ... I really would like to deal for some three card monte. I don't see why. My Bayesian probabilities would tell me not to bet money (because I've seen evidence that you would cheat). I was not born with this belief, but having seen lots of evidence for it, hold it fairly strongly, and would only be convinced otherwise by sailing over the edge. Show me the edge, and I'll give you $100, but forgive me for asking to hold your $1 in the meantime. There are 'non-round' solutions without edges. Ok, or some equally strong evidence. You *do* consider it highly probable that the world is round, right? As described above, to be 'convinced' you used a Bayesian prior of 1.0. You are not allowing for your (or our) ignorance. You keep saying that, but it isn't so. My Bayesian prior for the big bang was about 0.05. Ralph Hartley. |
#15
|
|||
|
|||
Galaxies without dark matter halos?
greywolf42 wrote:
Ralph Hartley wrote: greywolf42 wrote: .... The Bayesian probability is whatever you set the prior. Modified by the evidence you have seen so far. If you pick a good prior, the evidence will eventually win over your prejudices. The key word being "eventually." Here there is only one "data" point -- the prior. Which is why Bayesian statistics are invalid here. Huh? Has all astronomy in the past century not provided any data at all? Before I heard of (I wasn't born when it was discovered) the Hubble law I would have set the probability of the BB (if I had thought of it at all) to maybe P_prior(BB)~=0.05. That's my prior. Don't say it is 1. It is not. Until I read about the microwave background, I preferred the Steady State universe. Say P_hubble(BB)~=0.25. Even after that I wasn't completely convinced (their frequency coverage wasn't that good) P_cmb(BB)~=0.7. It only moved into "practically certain" territory when I saw the COBE data. Since then lots of data has come in. Light element abundances (I think this was known before COBE but I heard of it later), gravitational lensing, apparent evolution (distant galaxies etc. look different), CMB anisotropy, etc.. All of which have enough parameters that they don't' *have* to match *any* Big Bang variant, but they *do*. I stop keeping track at 99%. Do you actually *know* of a non-BB theory that predicts the black body CMB? What fluctuations does it predict? What element abundances? What red shifts? .... The primary difference for the BB is the difference between non-baryonic 'dark matter' (required by the BB to keep the light elements 'cooked' right) and just as-yet-unseen-because-it's-not-highly-luminous normal matter (sometimes confusingly called dark matter) that would be used by non-BB theories. Baryonic or not, you have to explain why some galaxies have it and some (assuming, for the sake of argument, that the report was correct) do not. .... NON-Bayesian probability doesn't apply *at* *all* to unique events, so why would I want to do such a thing? We are applying probability to cosmological *theories* -- not to a theoretical event contained within only SOME of those theories. If I use the other theories, then the Bayesian probability of the BB event is zero. It doesn't matter what theory you *use*, only which is *right*. There is no meaningful way to apply non-Bayesian probability to theories, since only one is right, and it's always the same one. If someone says "probability of the Big Bang" they are either talking nonsense, or talking about Bayesian probability. When I assumed you were talking about Bayesian probability I was giving you the benefit of the doubt, maybe that was wrong:-). Again, we can compare any two theories by relative probability (this does not require Bayesian probability). It does, even if you don't admit it. And we can arbitrarily select the 'most probable' theory and call it a '1.0' probability. We could, but that would be wrong, and I never did that. The probability of the best theory might not be even close to 1. Especially if there is a continuum of possible theories (there usually is). A set of theories (e.g. the set of BB theories) can have a combined probability near 1 even if every member of the set has infinitesimal probability. A good Bayesian never sets anything to 1. I said it was *close* to 1. ... One cannot invoke a Bayesian prior (which is an arbitrary choice) and then use this arbitrary choice to claim an event within the theories is a 'near certainty.' Of course I don't use my prior of 0.05 to claim that it is a "near certainty". I use my posterior of 0.99+. .... Do I get to define 'more or less round?' Within reasonable limits. If you can't guess approximately what I would consider reasonable, there is not much point in arguing. Tell the truth, you *do* know what I mean when I say "the earth is round", don't you? ... I really would like to deal for some three card monte. I don't see why. My Bayesian probabilities would tell me not to bet money (because I've seen evidence that you would cheat). I was not born with this belief, but having seen lots of evidence for it, hold it fairly strongly, and would only be convinced otherwise by sailing over the edge. Show me the edge, and I'll give you $100, but forgive me for asking to hold your $1 in the meantime. There are 'non-round' solutions without edges. Ok, or some equally strong evidence. You *do* consider it highly probable that the world is round, right? As described above, to be 'convinced' you used a Bayesian prior of 1.0. You are not allowing for your (or our) ignorance. You keep saying that, but it isn't so. My Bayesian prior for the big bang was about 0.05. Ralph Hartley. |
|
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
"Dark matter" forms dense clumps in ghost universe (Forwarded) | Andrew Yee | Astronomy Misc | 0 | November 21st 03 04:41 PM |
Galaxies without dark matter halos? | greywolf42 | Astronomy Misc | 34 | November 5th 03 12:34 PM |
A Detailed Map of Dark Matter in a Galactic Cluster Reveals How Giant Cosmic Structures Formed | Ron Baalke | Astronomy Misc | 3 | August 5th 03 02:16 PM |
Galaxies without dark matter halos? | Ed Keane III | Research | 4 | August 4th 03 12:39 PM |
Hubble tracks down a galaxy cluster's dark matter (Forwarded) | Andrew Yee | Astronomy Misc | 0 | July 17th 03 01:42 PM |