You are definitely punching above my weight class here but what the heck.
I do remember learning once that information theory states that the Gaussian distribution contains the minimum amount of information that any distribution can. Can't say why or even precisely what that means but I do recall people using that bit of knowledge to justify the *assumption* that a variable is Gaussian in the absence of better information.
The thinking being that such an assumption injects the minimum amount of information into the system which seems to be the right thing but nonetheless that alone doesn't save you from making a huge mistake.
I have some experience with Weibull and therefore I know it's derivation comes from (in Weibull's own seminal paper) a chain-of-links analogy. In this instance if the strengths of the individual links are not independent quantities then your derivation breaks down and you should be assuming something else. (In this case Weibull has chameleon-like properties so if you can estimate the distribution from sample data then you would probably be ok unless the lower tail is important because Weibull is lower-bounded.)
The problem seems to be finding sample datasets to test your assumptions about parameter distributions. As well things like the banking meltdown are extreme cases in the statistical sense so the properties of your distributions in the tail become critical and they are notoriously painful to estimate...meaning you need lots and lots of good data.
I think what these analysts are trying to do is a nightmare just given my general experience in an unrelated field with numerical methods and stats. In the real world both are, as you so aptly put, a pain in the ass.
I was surprised to find out some things that the President and other people in the US government don't know. GW Bush and Bill Clinton had a "debate" last week in Toronto and when someone asked about the new requirement to have a passport to cross the Canadian border (oops sorry, the "Northern" border) they were both taken aback. Clinton ok maybe but GW? His government instigated this for Pete's sake.
Also recently Janet Napolitano the Sec of Homeland Security claimed that the 9/11 hijackers entered the US via Canada. False. No single US/Canada issue makes Canadians more wild with frustration than this one and the freaking Sec of HS doesn't freaking know the truth. Anyways.
The point I'm trying to make is that yes it's quite possible Obama knows nothing about what's going on with copyright law and his government's role in it.
Didn't mean to trespass on anyone's specialty but the OP set the context of a enthusiast or amateur who is trying to do financial modeling. Isn't it better to get a quantitative statement that "my model is next to useless" rather than just a gut feel? I think so. Monte Carlo is a realistic way for a non-statistician to do that. Yes ok the analyst could also take partial derivatives analytically or numerically and apply a transformation to the input distribution and combine the variances to get some kind of output but frankly that's beyond most people who aren't specialists.
Also I didn't say anything about the validity of the underlying model. I was talking about a sensitivity analysis which can give one an idea of how serious a problem a poorly quantified input parameter really is. Economists who are better men that me have stated that trying to model human behaviour is futile. Sorry Foundation Trilogy. But maybe sometimes some models sort of work in limited ranges of application.
HaHAHAHA.
Jeez. Good thing I didn't mention Waloddi Weibull.
...you always reach a point where there's this one number that is completely made up...
***Try a sensitivity analysis using Monte Carlo techniques. That sounds hard but it isn't. Take the parameter that you have doubts about and give it a distribution (Gaussian or rectangular or something) with the mean at your best guess and the std deviation chosen to be big enough to cover the range it might reasonably vary over.
***Use Gaussian if you have an idea of what the parameter probably is but aren't exactly sure, rectangular if you really have no idea. A rectangular distribution says "I have no idea, the parameter could be anywhere within this particular range."
***Then run your analysis a million times with the parameter selected randomly from it's distribution for each run.
***That gives you a stochastic dataset of results. You can run simple stats on that data set to find the mean and std dev of the result value. You will then know how sensitive your results are to your poorly understood input parameter. If your 2-sigma output tells you the expected rate of return on a particular investment varies between -50% to +50% then you will know your model is pretty much useless and you will be doing a better job than the vast majority of professional analysts.
****Monte Carlo is great for those of us who don't care to learn the arcane minutiae of stat math. If you have a working model it takes an hour or two to extend it so you get stochastic results. Note that it's no harder to give a distribution to all your input parameters not just one. In which case you will be doing the kind of work that people who make 500 grand a year do.
Talking about things that are not my business is my business. And business is gooooood.
Also just because something is not stealing doesn't mean that it isn't not the opposite of stealing. By which I mean, inconspicuous recidivism. On the other hand, I forgot what I was talking about.
Bear that in mind the next time the dry cleaner ruins your floral drapes.
If a Newfoundland fisherman harpoons himself in the foot during next year's seal hunt they can give him one of these in the hospital to cheer him up.
It's a robot right? Maybe they could put missiles on (in?) one and put in on an ice flow then film it shooting back at the fisherdudes. That would be a YouTube classic right there boy I'll tell you what.
But lost behind it all is the primary problem â" "Thou shalt not steal".
The primary problem is not that people are stealing, the primary problem is that people don't think they are stealing.
And the primary question is: is the problem a problem with the moral health of people, or is the problem a problem with the entertainment industry's business model?
Are people as a collective allowed to decide what is publicly transferable? I would say, yeah. That's a bummer for those who profit when copies of works are scarce in the economic sense but then again times change. And the Ten Commandments don't contain any guarantees from God about the minimum level of profitability of the music business.
Of course one should always obey the laws of the land. Except when one shouldn't. For example, civil disobedience in protest of the arbitrary and disproportionate victimization of ordinary people by powerful elites has always gotten sympathetic treatment in the history books.
On this one, I predict the history books will portray the industry as a callous group who tried to enforce their will on the populace by making people terrified of their wrath.
The cinc tag (or the way I originally learned it "correlation doesn't necessarily imply causation") is overused for my tastes regardless of it's applicability. It's a bland truism.
Think of it this way, cinc could be replaced by "correlation isn't not causation" and be just as true.
An inflection is where the 2nd derivative of a function changes sign ie the curvature is zero.
I think the summary meant minima, that's where the first derivative is zero if the curve is smooth. That would mean it changes from going down to going up.
Unless we are talking about AMD's rate of change of growth which could go from shrinking faster and faster to shrinking slower and slower at an inflection. I guess that could be seen as a good thing these days for them.
Always remember AMD measures power consumption differently from Intel, AMD's number is more like a max envelope value while Intel's numbers are a type of average.
These guys found that a 45nm engineering Phenom II drew 24W less than a 9600 under load.
We have both Phenoms and C2Qs at work and I don't notice any difference in the noise. Maybe I would at home where it's quieter.
I always thought it would be brilliant if the Democrats developed a policy platform based on competition. Real competition.
Where the vision would be a marketplace where the small guy could take down the big guy based on brains and good ideas. The only tool the big guy would have to fight back would would be brains. Not legal shenanigans based on deep pockets, old boys clubs and family fortunes.
The policy should proudly proclaim that today's underprivileged are encouraged to drive today's upper class back to the middle class and trade places with them.
Because, in Western society, the upper classes are in grave danger of starting to consider themselves royalty.
That would completely outflank the Republicans claim to be the pro-business party leaving them with only the faith communities as a support base. Unfortunately the Democrats have trouble organizing anything more complicated than a birthday party.
Interesting to see how the paragons of capitalism don't believe in the free market.
A company I once worked for once had a written policy that anyone who had ever worked as a direct employee could not be hired at a later date as a contractor (contracting is very lucrative in this industry). I always thought that sounded legally dubious but despite some efforts the media had no interest in pursuing it.
I eventually left that company to contract at a competitor. On my last day the director of engineering told me "You realize I can't approve of this." To which I did not reply, but always wished I had "I can not approve of the way you accept public subsidies and then exported my job to Ireland."
Can't wait until I get a little older so I can name names.
Bilski dealt with business methods and not software so the "machine" part of the machine/transformation test never came up. The method failed the transformation test because the data didn't represent anything physical.
I find that an ominous development for software patents. The transformation part of the test is unavailable in the general case. The validity of software patents seem to rest solely on the fact that software runs on machines. Whether or not a computer suffices as a machine as patent law defines it is undecided. Given the restrictive definition applied to a patentable transformation, I'd say that future decisions will restrict how easy it is for software to invoke the computer to achieve patentability.
Actually, the Wikipedia page for Bilski quotes the court as having said:
"We leave to future cases the elaboration of the precise contours of machine implementation, as well as the answers to particular questions, such as whether or when recitation of a computer suffices to tie a process claim to a particular machine."
Slashdot Mirror
You are definitely punching above my weight class here but what the heck.
I do remember learning once that information theory states that the Gaussian distribution contains the minimum amount of information that any distribution can. Can't say why or even precisely what that means but I do recall people using that bit of knowledge to justify the *assumption* that a variable is Gaussian in the absence of better information.
The thinking being that such an assumption injects the minimum amount of information into the system which seems to be the right thing but nonetheless that alone doesn't save you from making a huge mistake.
I have some experience with Weibull and therefore I know it's derivation comes from (in Weibull's own seminal paper) a chain-of-links analogy. In this instance if the strengths of the individual links are not independent quantities then your derivation breaks down and you should be assuming something else. (In this case Weibull has chameleon-like properties so if you can estimate the distribution from sample data then you would probably be ok unless the lower tail is important because Weibull is lower-bounded.)
The problem seems to be finding sample datasets to test your assumptions about parameter distributions. As well things like the banking meltdown are extreme cases in the statistical sense so the properties of your distributions in the tail become critical and they are notoriously painful to estimate...meaning you need lots and lots of good data.
I think what these analysts are trying to do is a nightmare just given my general experience in an unrelated field with numerical methods and stats. In the real world both are, as you so aptly put, a pain in the ass.
I was surprised to find out some things that the President and other people in the US government don't know. GW Bush and Bill Clinton had a "debate" last week in Toronto and when someone asked about the new requirement to have a passport to cross the Canadian border (oops sorry, the "Northern" border) they were both taken aback. Clinton ok maybe but GW? His government instigated this for Pete's sake.
Also recently Janet Napolitano the Sec of Homeland Security claimed that the 9/11 hijackers entered the US via Canada. False. No single US/Canada issue makes Canadians more wild with frustration than this one and the freaking Sec of HS doesn't freaking know the truth. Anyways.
The point I'm trying to make is that yes it's quite possible Obama knows nothing about what's going on with copyright law and his government's role in it.
Peace.
Oh dear! Territoriality!
Didn't mean to trespass on anyone's specialty but the OP set the context of a enthusiast or amateur who is trying to do financial modeling. Isn't it better to get a quantitative statement that "my model is next to useless" rather than just a gut feel? I think so. Monte Carlo is a realistic way for a non-statistician to do that. Yes ok the analyst could also take partial derivatives analytically or numerically and apply a transformation to the input distribution and combine the variances to get some kind of output but frankly that's beyond most people who aren't specialists.
Also I didn't say anything about the validity of the underlying model. I was talking about a sensitivity analysis which can give one an idea of how serious a problem a poorly quantified input parameter really is. Economists who are better men that me have stated that trying to model human behaviour is futile. Sorry Foundation Trilogy. But maybe sometimes some models sort of work in limited ranges of application.
HaHAHAHA.
Jeez. Good thing I didn't mention Waloddi Weibull.
Can I quote that next time I teach?
No you may not Mr. Snarky Pants.
...you always reach a point where there's this one number that is completely made up...
***Try a sensitivity analysis using Monte Carlo techniques. That sounds hard but it isn't. Take the parameter that you have doubts about and give it a distribution (Gaussian or rectangular or something) with the mean at your best guess and the std deviation chosen to be big enough to cover the range it might reasonably vary over.
***Use Gaussian if you have an idea of what the parameter probably is but aren't exactly sure, rectangular if you really have no idea. A rectangular distribution says "I have no idea, the parameter could be anywhere within this particular range."
***Then run your analysis a million times with the parameter selected randomly from it's distribution for each run.
***That gives you a stochastic dataset of results. You can run simple stats on that data set to find the mean and std dev of the result value. You will then know how sensitive your results are to your poorly understood input parameter. If your 2-sigma output tells you the expected rate of return on a particular investment varies between -50% to +50% then you will know your model is pretty much useless and you will be doing a better job than the vast majority of professional analysts.
****Monte Carlo is great for those of us who don't care to learn the arcane minutiae of stat math. If you have a working model it takes an hour or two to extend it so you get stochastic results. Note that it's no harder to give a distribution to all your input parameters not just one. In which case you will be doing the kind of work that people who make 500 grand a year do.
EOM
But are the fluorescent monkeys beige?
...you really have no business talking about it.
Talking about things that are not my business is my business. And business is gooooood.
Also just because something is not stealing doesn't mean that it isn't not the opposite of stealing. By which I mean, inconspicuous recidivism. On the other hand, I forgot what I was talking about.
Bear that in mind the next time the dry cleaner ruins your floral drapes.
If a Newfoundland fisherman harpoons himself in the foot during next year's seal hunt they can give him one of these in the hospital to cheer him up.
It's a robot right? Maybe they could put missiles on (in?) one and put in on an ice flow then film it shooting back at the fisherdudes. That would be a YouTube classic right there boy I'll tell you what.
But lost behind it all is the primary problem â" "Thou shalt not steal".
The primary problem is not that people are stealing, the primary problem is that people don't think they are stealing.
And the primary question is: is the problem a problem with the moral health of people, or is the problem a problem with the entertainment industry's business model?
Are people as a collective allowed to decide what is publicly transferable? I would say, yeah. That's a bummer for those who profit when copies of works are scarce in the economic sense but then again times change. And the Ten Commandments don't contain any guarantees from God about the minimum level of profitability of the music business.
Of course one should always obey the laws of the land. Except when one shouldn't. For example, civil disobedience in protest of the arbitrary and disproportionate victimization of ordinary people by powerful elites has always gotten sympathetic treatment in the history books.
On this one, I predict the history books will portray the industry as a callous group who tried to enforce their will on the populace by making people terrified of their wrath.
Make sure their lawyers are disbarred too.
Especially the five that now hold senior positions at the Department of Justice.
Is TPB TPTB or TPHB?
Hey Thai King, is that a hat or the nose cone of a North Korean rocket that landed on your head?
The Thai King's mother is so fat when she gets in an elevator it goes down.
So...who's going to register protecttehking.net and put up an nice pie-in-the-king's-face flash game?
Don't look at me I'm the brains of this operation.
The cinc tag (or the way I originally learned it "correlation doesn't necessarily imply causation") is overused for my tastes regardless of it's applicability. It's a bland truism.
Think of it this way, cinc could be replaced by "correlation isn't not causation" and be just as true.
Where will this end? First they'll make a cellphone that has no eyes and the next thing you know they'll make one with no spleen.
I better hold onto my Nokia SpleenMaster 5000, it might be the last of its kind.
This may be an inflection point for AMD...
An inflection is where the 2nd derivative of a function changes sign ie the curvature is zero.
I think the summary meant minima, that's where the first derivative is zero if the curve is smooth. That would mean it changes from going down to going up.
Unless we are talking about AMD's rate of change of growth which could go from shrinking faster and faster to shrinking slower and slower at an inflection. I guess that could be seen as a good thing these days for them.
Slow day I'll go away now.
Perhaps this might explain something:
https://login.yahoo.com/config/login_verify2?.partner=vz-acs&.done=http%3a//verizon.yahoo.com
Verizon and Yahoo have some sort of integrated portal.
Non-story.
Always remember AMD measures power consumption differently from Intel, AMD's number is more like a max envelope value while Intel's numbers are a type of average.
These guys found that a 45nm engineering Phenom II drew 24W less than a 9600 under load.
We have both Phenoms and C2Qs at work and I don't notice any difference in the noise. Maybe I would at home where it's quieter.
I always thought it would be brilliant if the Democrats developed a policy platform based on competition. Real competition.
Where the vision would be a marketplace where the small guy could take down the big guy based on brains and good ideas. The only tool the big guy would have to fight back would would be brains. Not legal shenanigans based on deep pockets, old boys clubs and family fortunes.
The policy should proudly proclaim that today's underprivileged are encouraged to drive today's upper class back to the middle class and trade places with them.
Because, in Western society, the upper classes are in grave danger of starting to consider themselves royalty.
That would completely outflank the Republicans claim to be the pro-business party leaving them with only the faith communities as a support base. Unfortunately the Democrats have trouble organizing anything more complicated than a birthday party.
Interesting to see how the paragons of capitalism don't believe in the free market.
A company I once worked for once had a written policy that anyone who had ever worked as a direct employee could not be hired at a later date as a contractor (contracting is very lucrative in this industry). I always thought that sounded legally dubious but despite some efforts the media had no interest in pursuing it.
I eventually left that company to contract at a competitor. On my last day the director of engineering told me "You realize I can't approve of this." To which I did not reply, but always wished I had "I can not approve of the way you accept public subsidies and then exported my job to Ireland."
Can't wait until I get a little older so I can name names.
Bilski dealt with business methods and not software so the "machine" part of the machine/transformation test never came up. The method failed the transformation test because the data didn't represent anything physical.
I find that an ominous development for software patents. The transformation part of the test is unavailable in the general case. The validity of software patents seem to rest solely on the fact that software runs on machines. Whether or not a computer suffices as a machine as patent law defines it is undecided. Given the restrictive definition applied to a patentable transformation, I'd say that future decisions will restrict how easy it is for software to invoke the computer to achieve patentability.
Actually, the Wikipedia page for Bilski quotes the court as having said:
"We leave to future cases the elaboration of the precise contours of machine implementation, as well as the answers to particular questions, such as whether or when recitation of a computer suffices to tie a process claim to a particular machine."
So I think the question remains open.
Hey, since that is a transformative act, why not patent it?
Sorry, prior art. Look up Genghis Khan v Asia.