I've read the comments in the source code. I see nothing that is incompatible with the kind of sensitivity analysis and exploratory data analysis that is legitimately and routinely done in many contexts.
The original data has not been thrown away. The original data is retained by the meteorological services that acquired them, not by CRU.
It is also a myth that the global warming theory is inconsistent with temperatures over the last 10 years. In fact, temperatures over the last 10 years are well within the range of variation seen in runs of global climate models. See here for detailed analysis by an expert statistician. By the way, the graphs presented there show many examples of natural variation producing warming above the model projection (in 1998, for example), refuting your claim that "people claim there's just a natural variation which has been cooling the planet, but those people claim every warming is due to CO2 and not natural variation." In fact, excursions of measured temperature above the projected long-term trend are as frequent, and as large, as the excursions below the trend.
Some positive feedbacks are unavoidable based on the basic physics. For example CO2 produces warming, which increases water vapor, which further warms the planet, which reduces solubility of CO2, causing further CO2 release. This is a positive feedback. So far, nobody has been able to find negative feedbacks sufficient to overwhelm the known positive feedbacks, nor has anybody been able to get a model without net positive feedbacks to come anywhere close to being consistent with the historical record of climate.
The role of CO2 in warming the earth was first worked out by Arrhenius in the 1900's. A nice summary of the development of scientific thought regarding CO2 and climate modeling can be found here
No, most of this data has been available for years, although RealClimate assembled this convenient index in part as a response to claims that scientists were withholding data that were leveled in association with the recent data theft from CRU.
There has long been plenty of raw and corrected climate data, as well as climate models with source code, more than enough for any interested investigator to replicate major conclusions of climate researchers.
Most of the interest in raw data seems to be politically motivated and to be directed toward finding pretexts to level accusations against climate researchers in order to create the illusion of doubt about the science. The groups that scream most loudly about unreleased data never seem to do much with the great bulk of data that is available. One popular strategy seems to be to demand raw data from somebody who does not actually own it, and then cry "conspiracy" when they try to explain that unreleased raw data must be requested from the actual owner. For example CRU was bombarded with "freedom of information" demands for raw data that was not generated by them and that actually belonged to national meteorological services.
The "greenhouse gas theory" came into being when the Mauna Loa observations began to show the increase in the atmospheric co2 concentration and the climate was generally on a warming cycle.
This is nonsense. The role of CO2 in warming the planet has been known for a century, and is an unavoidable consequence of radiative physics. With greater computer power and development of more detailed models, physical climate models have progressively gotten better over the years in calculating the impact of CO2 and other "greenhouse gasses" on the radiative balance of the earth.
Hardly. This looks to me like the kind of sensitivity analysis that any good laboratory does frequently. "I think that the decline in this data might be an artifact" "Well, how much do we need to worry about it? How big an impact is it having on conclusions?" "OK, I'll take a guess at how much it is in error, apply a correction, and see how much it changes the output."
Such code is never meant for publication, but just to help scientists understand their data and as a guide for future research. It is typical that a big all-caps warning "APPLY ARTIFICIAL CORRECTION" comment is included, so that nobody will get confused and accidentally include this "what-if" analysis in a publication. And indeed, nobody has been able to identify any actual publication in which there are undocumented "ARTIFICIAL CORRECTIONs"
With the global warming 'scandal', you have a few scientists who are the only ones with access to the raw temperature data.
This is a myth that is being propounded for political purposes. In reality, climate research is one of the most open areas of scientific study. I don't know of any other field in which so much of the raw data and even the analysis software has been made publicly available. A huge amount of data and computer code, including a great deal of raw climate data is available online. There is some data that is owned by various national meteorological services and not publicly available, but even this is routinely provided to qualified researchers. The conclusions of CRU (the owner of the stolen files) have been replicated by multiple independent research groups that have carried out their own independent analyses of the same data.
The emails and SOURCE CODE are authentic and are absolute proof of the bullshit going on. Go read through it. It's all there in 1s and 0s. No amount of "context" can change "we threw away data and created bogus graphs to get our desired result" into "we're just misunderstood scientists trying to save the world".
It is common practice in scientific work to test for the impact of errors or biases in particular datasets by substituting artificial data for real data in order to determine whether the conclusions are sensitive to a particular type of error or bias. Such "sensitivity analyses" are done to help researchers understand how robust their conclusions are, and to estimate the potential impact of possible errors in the data. This is perfectly legitimate, and good scientific practice. So finding some source code in which artificial data is substituted for real is not evidence of anything dishonest, unless you can show that the results were published and misrepresented as actual data. Yet weeks after the release of the stolen files, nobody has managed to identify any publication that contains fabricated data.
So it is fraudulent to use direct measurement of actual temperatures in place of an indirect temperature measurement when other factors impair the accuracy of the indirect measurement--and to document that choice and the scientific basis for it in great detail in the peer-reviewed scientific literature?
No, the source code that was leaked wasn't climate model code. That's not what CRU does. The guys who analyze historical trends aren't generally the guys who create climate models. Different skill sets; trend analysis is statistics, climate modeling is physics. Besides, the source code for the climate models is available online (and has been for a long time). Anybody who wants can work with it.
In fact, it's not clear whether the code that was leaked was ever used for any published data. Anybody who has done research probably has code on their HD where they've asked questions like, "Well, what if I assume that this data is wrong and make this modification to the data. Does that change my conclusions?" A lot of the time, the result is not particularly interesting, and it never gets published.
The real problem is that matching the wiring diagram of a simulated brain to that of a cat probably will not result in a computer that thinks like a cat, because a brain is a dynamic entity. All sorts of things change with brain activity, including numbers of receptors, their properties, their distribution, the shape of synaptic spines, etc., etc. We don't know how much of this is really critical to the major functions of the brain, and how much of this is minor evolutionary tweaks, or even fortuitous "spandrels" that are present for incidental reasons but are not crucial to function. So perhaps a simple rule for how synaptic coupling changes over time will suffice, or perhaps it is necessary to carry the simulation down to the level of individual receptors moving around on the cell membrane.
So simply setting out to simulate a relatively large mammal like a cat is pretty much a waste of time, because it is virtually certain that the first attempt won't begin to reproduce the function of the animal brain. A model would be very useful for hypothesis testing, but it makes sense to start small. I'd be happy to see a simulated mouse brain that could do simple things like finding its way around a room or remembering where food is hidden. Even that is probably too ambitious to start with. How about a simulated fly or cockroach that reproduces the behavior (in a simulated space) of the actual insect?
I think Apple very much cares. I think they want to keep the Mac market locked up for themselves and avoid the mistakes of IBM in the 80s.
I think that the reason Apple doesn't care about individual users running hacked versions of OS X on non-Macs is that it does not really threaten either of those goals. Aside from a small minority of hackers, most people don't want to deal with the potential aggravation of running an operating system on an unsupported platform. You can't simply install upgrades and patches--you have to go on the internet and check on whether the upgrades are compatible, and wait to download modified versions if they aren't. And if your software doesn't work right on your Hackintosh, who do you go to for help? Not the local Apple store, obviously. And a manufacturer of a program intended to run on a real Mac probably will not have much help to offer either. So you are dependent upon the Hackintosh community. Computer enthusiasts thrive on this sort of thing, but most of Apple's customers simply don't want to be bothered.
You are right, I am forgetting about the Altair and the machines that followed it. These, too, demonstrate the point that open hardware standards are a good thing. I think the open standards used within the PC are of enormous importance as an explanation of the PC's success and its market dominance over technically superior but proprietary machines.
It seems hard to reconcile the notion that open standards are critical for success with the fact that the early open standards systems--the S100, CP/M computers--never achieved great success despite a substantial lead in the marketplace, and ultimately fell by the wayside and were supplanted by closed (or semi-closed) systems: the Apple II, Commodore 64, Atari ST, and IBM PC (which was originally a closed system due to its proprietary BIOS).
It seems that a stronger argument can be made that closed standards are essential to commercial success, but that once a closed standard-system achieves a substantial measure of success, it creates opportunities for open-standard systems to exist in parallel.
I would speculate that Apple is not really threatened by Frankenmacs in general. The kind of hobbyist who is technically inclined and is willing to put something like this together is probably outside of their target audience. I have known people who bought Macs not because they were fans of Apple, but because they were dissatisfied with PCs loaded with Windows.
No, I don't think Apple much cares, or at least they realize that suing users is a losing proposition. Apple's protection strategy is little more than the equivalent of a chain on the door. It doesn't so much keep people out as place them in the position of having to break something to get in, thereby serving notice that people who run Mac OS X on foreign hardware are in violation of their user agreements. Of course, the chain is cut almost immediately, as Apple surely knows it will be. They just put it back with each new release, reminding everybody, "this upgrade is for our computers only, and if you want to violate our rights, you'll have to take the trouble to patch it. '
Before then, there were home computers, but all of them had proprietary designs and clones were effectively illegal.
This is nonsense. There were plenty of generic S-100 computers that ran CP/M (of which MS-DOS was a rebranded clone, one that likely would would be in violation of today's copyright laws). Apple was one of several companies (Commodore, Tandy, Exidy) to provide an all-in-one PC design. The Exidy PC even used the S-100 standard. But Apple's design offered the most for the price, providing color graphics (albeit limited) on a consumer TV out of the box, as well true bitmapped graphics (and eventually, a low-cost floppy drive). And while proprietary, the Apple II was "open" in a number of respects--it came with a listing of its ROM and standards for interface cards.
The PC's success probably owes more to IBM's business reputation than to any great virtues of the hardware or software design (which were adequate, but hardly impressive compared to its competitors), or to its eventual "openness." Indeed, I'd argue that the opening up of the platform was a consequence of its success rather than the cause--the commercial success of the platform made it a worthwhile gamble for other companies to invest in cloning its ROM.
Random mutations haven't yet given our species overproduction of this receptor protein. I'm going out on a limb here and assuming that they got over-expression by repeating the gene sequence. Tandem duplications usually happen in specific places in the genome where the DNA replication enzymes get "stuck" when replicating the parent strand, causing repeats.
This is the most convenient way to do it for experimental purposes, but it is far from the only way that the level of a protein can be increased. This can happen by increased activity of the promoter; increased levels, binding, or phosphorylation of transcription factors; increased translation; decreased inhibition by small RNAs; increased trafficking to the surface; decreased degradation; decreased activity of regulatory factors that mediate exocytosis, etc., etc. So it really is not very plausible that there is zero genetic variation in the population above the average expression level of the protein.
Over-expression of this protein doesn't give any kind of fitness
However, it is hard to imagine that a rodent wouldn't reap a Darwinian benefit (i.e. increased reproduction or survival of offspring) from being better able to remember the location of food, possible hazards, etc. Unless, of course, the behavioral tests of intelligence that are used in the laboratory do not accurately reflect the intellectual demands of life in the wild.
There is an upper limit on intelligence for breeding. I've known plenty of smart guys and gals that felt like reproduction was a waste of their time, so perhaps our breeding process selects against high IQ's.
This also seems like a very plausible hypothesis, but kind of vague. I was wondering what particular behavioral traits might be altered to produce a decrement in reproductive success sufficient to balance the likely advantage of better learning and memory.
Except that it also seems rather implausible that there would be no selective advantage to remembering better than the other rats how to get to the food, unless there is some other negative.
Selection pressure works in the negative. Even if it adds a benefit, the individual critters without it would have to fail to breed. As long as the current is sufficient, we'll get more of the current.
This is a misconception. Nobody necessarily has to fail to breed. All that is required for a mutation to increase in the population is for animals with a mutation to have more offspring, or more that survive to breed, than the average.
Or (IMHO) the most probable presumption is that rats are not at the peak of their evolution and this particular trait just hasn't occurred yet or would not be a naturally occurring mutation.
Except that overexpression mutations tend to be common. There are just so many ways in which an existing protein can be upregulated: Increased transcription, increased translation, increased membrane insertion, decreased degradation, changes in cellular feedback regulation, etc., etc. So it is highly implausible to suppose that there is no existing genetic variation in levels of NR2B for selection to act upon. Given the short generation span of rodents, they should be pretty close to the local optimum of expression for all existing proteins.
Another possibility is that rats and mice in the wild already have this, but that research animals, which have been through many generations of selection for docility and manageability, under-express NR2B.
Basically, a rat only has to be as smart as necessary to reproduce. Perhaps there is no advantage to a longer memory, remembering longer simply doesn't get you more tail.
However, you need to survive to reproduce. Rats use their intelligence to get access to, and remember the location of, food sources. So it seems unlikely that a rat that was smarter, with no downside, would not have a selective advantage when food is scarce.
It HAS happened, but those affected (rats included) simply can't get laid to propagate the phenomenon...
I expect that you were joking, but I think that it is a serious possibility. Transgenic animals are bred in captivity. I wonder what their mating success would be in the wild.
There is a third possibility, namely that the set of mutations necessary to give rise to this advantage are too improbable to occur (or perhaps even fundamentally impossible).
If it were something complicated, like multiple specific mutations or introduction of a gene that the animal doesn't have to begin with, this is plausible. But this is merely increasing the expression of a gene that is already there, and there are multiple regulatory mechanisms, both genomic and postgenomic that control its abundance on neurons. So it is difficult to imagine that there is not underlying variation in expression levels that natural selection could use to further optimize its expression.
Slashdot Mirror
I've read the comments in the source code. I see nothing that is incompatible with the kind of sensitivity analysis and exploratory data analysis that is legitimately and routinely done in many contexts.
The original data has not been thrown away. The original data is retained by the meteorological services that acquired them, not by CRU.
It is also a myth that the global warming theory is inconsistent with temperatures over the last 10 years. In fact, temperatures over the last 10 years are well within the range of variation seen in runs of global climate models. See here for detailed analysis by an expert statistician. By the way, the graphs presented there show many examples of natural variation producing warming above the model projection (in 1998, for example), refuting your claim that "people claim there's just a natural variation which has been cooling the planet, but those people claim every warming is due to CO2 and not natural variation." In fact, excursions of measured temperature above the projected long-term trend are as frequent, and as large, as the excursions below the trend.
Some positive feedbacks are unavoidable based on the basic physics. For example CO2 produces warming, which increases water vapor, which further warms the planet, which reduces solubility of CO2, causing further CO2 release. This is a positive feedback. So far, nobody has been able to find negative feedbacks sufficient to overwhelm the known positive feedbacks, nor has anybody been able to get a model without net positive feedbacks to come anywhere close to being consistent with the historical record of climate.
The role of CO2 in warming the earth was first worked out by Arrhenius in the 1900's. A nice summary of the development of scientific thought regarding CO2 and climate modeling can be found here
No, most of this data has been available for years, although RealClimate assembled this convenient index in part as a response to claims that scientists were withholding data that were leveled in association with the recent data theft from CRU.
There has long been plenty of raw and corrected climate data, as well as climate models with source code, more than enough for any interested investigator to replicate major conclusions of climate researchers.
Most of the interest in raw data seems to be politically motivated and to be directed toward finding pretexts to level accusations against climate researchers in order to create the illusion of doubt about the science. The groups that scream most loudly about unreleased data never seem to do much with the great bulk of data that is available. One popular strategy seems to be to demand raw data from somebody who does not actually own it, and then cry "conspiracy" when they try to explain that unreleased raw data must be requested from the actual owner. For example CRU was bombarded with "freedom of information" demands for raw data that was not generated by them and that actually belonged to national meteorological services.
This is nonsense. The role of CO2 in warming the planet has been known for a century, and is an unavoidable consequence of radiative physics. With greater computer power and development of more detailed models, physical climate models have progressively gotten better over the years in calculating the impact of CO2 and other "greenhouse gasses" on the radiative balance of the earth.
Hardly. This looks to me like the kind of sensitivity analysis that any good laboratory does frequently.
"I think that the decline in this data might be an artifact"
"Well, how much do we need to worry about it? How big an impact is it having on conclusions?"
"OK, I'll take a guess at how much it is in error, apply a correction, and see how much it changes the output."
Such code is never meant for publication, but just to help scientists understand their data and as a guide for future research. It is typical that a big all-caps warning "APPLY ARTIFICIAL CORRECTION" comment is included, so that nobody will get confused and accidentally include this "what-if" analysis in a publication. And indeed, nobody has been able to identify any actual publication in which there are undocumented "ARTIFICIAL CORRECTIONs"
This is a myth that is being propounded for political purposes. In reality, climate research is one of the most open areas of scientific study. I don't know of any other field in which so much of the raw data and even the analysis software has been made publicly available. A huge amount of data and computer code, including a great deal of raw climate data is available online. There is some data that is owned by various national meteorological services and not publicly available, but even this is routinely provided to qualified researchers. The conclusions of CRU (the owner of the stolen files) have been replicated by multiple independent research groups that have carried out their own independent analyses of the same data.
It is common practice in scientific work to test for the impact of errors or biases in particular datasets by substituting artificial data for real data in order to determine whether the conclusions are sensitive to a particular type of error or bias. Such "sensitivity analyses" are done to help researchers understand how robust their conclusions are, and to estimate the potential impact of possible errors in the data. This is perfectly legitimate, and good scientific practice. So finding some source code in which artificial data is substituted for real is not evidence of anything dishonest, unless you can show that the results were published and misrepresented as actual data. Yet weeks after the release of the stolen files, nobody has managed to identify any publication that contains fabricated data.
So it is fraudulent to use direct measurement of actual temperatures in place of an indirect temperature measurement when other factors impair the accuracy of the indirect measurement--and to document that choice and the scientific basis for it in great detail in the peer-reviewed scientific literature?
No, the source code that was leaked wasn't climate model code. That's not what CRU does. The guys who analyze historical trends aren't generally the guys who create climate models. Different skill sets; trend analysis is statistics, climate modeling is physics. Besides, the source code for the climate models is available online (and has been for a long time). Anybody who wants can work with it.
In fact, it's not clear whether the code that was leaked was ever used for any published data. Anybody who has done research probably has code on their HD where they've asked questions like, "Well, what if I assume that this data is wrong and make this modification to the data. Does that change my conclusions?" A lot of the time, the result is not particularly interesting, and it never gets published.
The real problem is that matching the wiring diagram of a simulated brain to that of a cat probably will not result in a computer that thinks like a cat, because a brain is a dynamic entity. All sorts of things change with brain activity, including numbers of receptors, their properties, their distribution, the shape of synaptic spines, etc., etc. We don't know how much of this is really critical to the major functions of the brain, and how much of this is minor evolutionary tweaks, or even fortuitous "spandrels" that are present for incidental reasons but are not crucial to function. So perhaps a simple rule for how synaptic coupling changes over time will suffice, or perhaps it is necessary to carry the simulation down to the level of individual receptors moving around on the cell membrane.
So simply setting out to simulate a relatively large mammal like a cat is pretty much a waste of time, because it is virtually certain that the first attempt won't begin to reproduce the function of the animal brain. A model would be very useful for hypothesis testing, but it makes sense to start small. I'd be happy to see a simulated mouse brain that could do simple things like finding its way around a room or remembering where food is hidden. Even that is probably too ambitious to start with. How about a simulated fly or cockroach that reproduces the behavior (in a simulated space) of the actual insect?
I think that the reason Apple doesn't care about individual users running hacked versions of OS X on non-Macs is that it does not really threaten either of those goals. Aside from a small minority of hackers, most people don't want to deal with the potential aggravation of running an operating system on an unsupported platform. You can't simply install upgrades and patches--you have to go on the internet and check on whether the upgrades are compatible, and wait to download modified versions if they aren't. And if your software doesn't work right on your Hackintosh, who do you go to for help? Not the local Apple store, obviously. And a manufacturer of a program intended to run on a real Mac probably will not have much help to offer either. So you are dependent upon the Hackintosh community. Computer enthusiasts thrive on this sort of thing, but most of Apple's customers simply don't want to be bothered.
It seems hard to reconcile the notion that open standards are critical for success with the fact that the early open standards systems--the S100, CP/M computers--never achieved great success despite a substantial lead in the marketplace, and ultimately fell by the wayside and were supplanted by closed (or semi-closed) systems: the Apple II, Commodore 64, Atari ST, and IBM PC (which was originally a closed system due to its proprietary BIOS).
It seems that a stronger argument can be made that closed standards are essential to commercial success, but that once a closed standard-system achieves a substantial measure of success, it creates opportunities for open-standard systems to exist in parallel.
No, I don't think Apple much cares, or at least they realize that suing users is a losing proposition. Apple's protection strategy is little more than the equivalent of a chain on the door. It doesn't so much keep people out as place them in the position of having to break something to get in, thereby serving notice that people who run Mac OS X on foreign hardware are in violation of their user agreements. Of course, the chain is cut almost immediately, as Apple surely knows it will be. They just put it back with each new release, reminding everybody, "this upgrade is for our computers only, and if you want to violate our rights, you'll have to take the trouble to patch it.
'
This is nonsense. There were plenty of generic S-100 computers that ran CP/M (of which MS-DOS was a rebranded clone, one that likely would would be in violation of today's copyright laws). Apple was one of several companies (Commodore, Tandy, Exidy) to provide an all-in-one PC design. The Exidy PC even used the S-100 standard. But Apple's design offered the most for the price, providing color graphics (albeit limited) on a consumer TV out of the box, as well true bitmapped graphics (and eventually, a low-cost floppy drive). And while proprietary, the Apple II was "open" in a number of respects--it came with a listing of its ROM and standards for interface cards.
The PC's success probably owes more to IBM's business reputation than to any great virtues of the hardware or software design (which were adequate, but hardly impressive compared to its competitors), or to its eventual "openness." Indeed, I'd argue that the opening up of the platform was a consequence of its success rather than the cause--the commercial success of the platform made it a worthwhile gamble for other companies to invest in cloning its ROM.
This is the most convenient way to do it for experimental purposes, but it is far from the only way that the level of a protein can be increased. This can happen by increased activity of the promoter; increased levels, binding, or phosphorylation of transcription factors; increased translation; decreased inhibition by small RNAs; increased trafficking to the surface; decreased degradation; decreased activity of regulatory factors that mediate exocytosis, etc., etc. So it really is not very plausible that there is zero genetic variation in the population above the average expression level of the protein.
However, it is hard to imagine that a rodent wouldn't reap a Darwinian benefit (i.e. increased reproduction or survival of offspring) from being better able to remember the location of food, possible hazards, etc. Unless, of course, the behavioral tests of intelligence that are used in the laboratory do not accurately reflect the intellectual demands of life in the wild.
This also seems like a very plausible hypothesis, but kind of vague. I was wondering what particular behavioral traits might be altered to produce a decrement in reproductive success sufficient to balance the likely advantage of better learning and memory.
Except that it also seems rather implausible that there would be no selective advantage to remembering better than the other rats how to get to the food, unless there is some other negative.
From the point of view of natural selection, having less children is a downside.
This is a misconception. Nobody necessarily has to fail to breed. All that is required for a mutation to increase in the population is for animals with a mutation to have more offspring, or more that survive to breed, than the average.
Except that overexpression mutations tend to be common. There are just so many ways in which an existing protein can be upregulated: Increased transcription, increased translation, increased membrane insertion, decreased degradation, changes in cellular feedback regulation, etc., etc. So it is highly implausible to suppose that there is no existing genetic variation in levels of NR2B for selection to act upon. Given the short generation span of rodents, they should be pretty close to the local optimum of expression for all existing proteins.
Another possibility is that rats and mice in the wild already have this, but that research animals, which have been through many generations of selection for docility and manageability, under-express NR2B.
However, you need to survive to reproduce. Rats use their intelligence to get access to, and remember the location of, food sources. So it seems unlikely that a rat that was smarter, with no downside, would not have a selective advantage when food is scarce.
I expect that you were joking, but I think that it is a serious possibility. Transgenic animals are bred in captivity. I wonder what their mating success would be in the wild.
If it were something complicated, like multiple specific mutations or introduction of a gene that the animal doesn't have to begin with, this is plausible. But this is merely increasing the expression of a gene that is already there, and there are multiple regulatory mechanisms, both genomic and postgenomic that control its abundance on neurons. So it is difficult to imagine that there is not underlying variation in expression levels that natural selection could use to further optimize its expression.
corrected link