[feedquote='E-Cat World','http://www.e-catworld.com/2016/10/20/rossi-quarkx-commercialization-depends-of-5-sigma/']We mentioned recently here that Andrea Rossi is focusing hard on obtaining a 5 Sigma rating for his Quark X reactor (5 Sigma in physics being a probability that an event happening the same way every time is 99.9999%). It seem now that there is a very practical reason for striving to reach this goal: […][/feedquote]
Rossi: QuarkX Commercialization Depends of 5 Sigma
-
Yes, sure, the partner now needs sigma 5 :dash:
Each investor seeing such an revolutionary technology wouldn't ask for high reliablity. It has to work. That's it.
Reliability and stability is something that can be engineered when enough money and skilled people are involved. -
Rossi has been adamant in explaining that his reactor does not use and does not produce radioactive material. This is required to avoid regulation and licensing of his reactors as a nuclear device by the authorities. But what Rossi and his backers do not yet understand is that Rossi's technology enriches nuclear material used in nuclear weapons.
There will come a time when all Rossi reactors will require licences from the authorities as a nuclear device. Rossi should clear this issue up before people invest in his technology and before production and fielding of his reactors begin. Rossi needs to be open and forthcoming about these issues and face this problem head on. :nuke:
-
Since 5 sigma relates to probability, meaning that there is approximately 1:3.5x106 chance that an observed effect is not genuine it doesn't necessarily indicate how large or small the signal is, just the likelihood that it is a real. I think it very unlikely that Rossi has performed 3.5M tests yet, so suspect that it is just a sciency-sounding buzzword he picked up from someone talking about reproducibility/repeatability of his latest invention.
-
I'm sure that the physicists borrowed 5 Sigma from the statisticians, so even calcualted as SD's it would still boil down to the same old 1 in 3.5 million. Timar, over on ECW asked if it was perhaps related to the number of cycles of operation. In which case if Q-X had a base frequency of say 1Hz then 3.5M cycles would take 40 days, not very good from an end-user point of view.
Without a crtierion showing what data 5 sigma is derived from (or applied to) it is all pretty meaningless. But still pretty.i
-
Timar, over on ECW asked if it was perhaps related to the number of cycles of operation. In which case if Q-X had a base frequency of say 1Hz then 3.5M cycles would take 40 days, not very good from an end-user point of view.
No, that wasn't me. I just noted that Rossi made it quite clear that it refers to reliability, not effect size and that it would of course require a timeframe to make any sense. -
Though it is right, that 5 sigma can be interpreted as an event, which occurrs 99,9999 the same way, physicists tend to look at this from the other side.
There is a likelhood of 0,000 057 3303%, that the event is not a error/statistical fluke or mismeasurement.
???? I think, he did not reach this value, but if any claims of Rossi would not be very near to this, why continue with ecat at all ???And from this 5 sigma, which we see all over the media even more since the discovery of the higgs boson, of course Rossi jumped on the train.
Perhaps he just wants to sound more professional with this, instead of his stupid F9, F8, whatever silly evaluations he used.But it is nothing more than another trick to keep his follower's attention, while he again, and still, is not able to deliver. What happend to that hotcat? No information, but now suddenly all refer to the quarkx.
This is nothing new, as this strategy worked for the hotcat, so he will just continue telling the same shit, with a new name, and now even with real, physical evaluation-factors.
What a pitty.
-
Considering the question of 5 sigma (and not Rossi), I am trying to understand the question of sample size, so someone who knows more about this please correct what is wrong with the following understanding: "5 sigma" refers to five standard deviations from the mean. If you have enough data for statistical significance, 1000 trials, as an example, then, and your mean and sd are 50 and +- 1, say, then any measurement above 55 or below 45 will be "5 sigma" away from the mean. Translating this to a Rossi-like experiment, if you have 1000 blank runs, that will give you a mean and standard deviation for the excess heat (or whatever figure of merit is being considered). And then if you run your active runs and show a measurement that is 5 standard deviations above the blank runs, then you have a "5 sigma" result.
Here we have been working with a sample size of order 1000, and not millions of runs. Can someone comment on sample size in this context?
-
Considering the question of 5 sigma (and not Rossi), I am trying to understand the question of sample size, so someone who knows more about this please correct what is wrong with the following understanding: "5 sigma" refers to five standard deviations from the mean. If you have enough data for statistical significance, 1000 trials, as an example, then, and your mean and sd are 50 and +- 1, say, then any measurement above 55 or below 45 will be "5 sigma" away from the mean. Translating this to a Rossi-like experiment, if you have 1000 blank runs, that will give you a mean and standard deviation for the excess heat (or whatever figure of merit is being considered). And then if you run your active runs and show a measurement that is 5 standard deviations above the blank runs, then you have a "5 sigma" result.
Here we have been working with a sample size of order 1000, and not millions of runs. Can someone comment on sample size in this context?
Eric,
I don't think you would look at one data point, and maybe you don't mean that. The 5 sigma would depend on a properly conducted calibration series and active run. It is easy to get 5 sigma in an invalid experiment.
There is a notion of statistical power that gives you an idea about the sample size needed to achieve a given level of statistical significance.
-
There is a notion of statistical power that gives you an idea about the sample size needed to achieve a given level of statistical significance.
Thank you, Jack, for the link to the description of statistical power. You make the good point that the number of active runs is important, and that a single result 5 sd above the (blank) mean wouldn't be adequate.- Can you think up a simple protocol that would show (or refute) "5 sigma," using the statistical power calculation you pointed to and giving concrete numbers of various runs as an illustration?
- Is your understanding that a 5 sigma result can be obtained in fewer than millions of blank/active runs? It seems to me that the ratio that is mentioned in connection with 5 sigma (e.g., 1 in 1744278) refers to the population size and not the sample size.
-
Quote
Considering the question of 5 sigma (and not Rossi), I am trying to understand the question of sample size, so someone who knows more about this please correct what is wrong with the following understanding: "5 sigma" refers to five standard deviations from the mean. If you have enough data for statistical significance, 1000 trials, as an example, then, and your mean and sd are 50 and +- 1, say, then any measurement above 55 or below 45 will be "5 sigma" away from the mean. Translating this to a Rossi-like experiment, if you have 1000 blank runs, that will give you a mean and standard deviation for the excess heat (or whatever figure of merit is being considered). And then if you run your active runs and show a measurement that is 5 standard deviations above the blank runs, then you have a "5 sigma" result.Here we have been working with a sample size of order 1000, and not millions of runs. Can someone comment on sample size in this context?
I can add a bit more to this. 5 sigma makes sense where data is stochastic distribution and inference from the data back to the hypothesis then depends on the amount of data corrected. 5 sigma would be the calculated likelihood of the data seen happening by chance, expressed as a distance along the normal distribution. There is technically a small difference between "two-sided sigma" and "one-sided sigma" but we can ignore that.
When the experiment generates clear data, but has a reliability problem, the actual model is different and best modelled as a mixture of two different stochastic models: e.g. what happens when the experiment works, or what happens when it does not. In more complex case this becomes a distribution of distributions where the first distribution determines "how much" the experiment is working. You might naively think there is no difference and all that matters is the overall probability but for hypothesis inference that different cases matter: so that a binary failure mode with a definite value for both cases is quite different from a stochastic value varying according to some normal distribution.
As always then things like 5 sigma can be applied correctly or incorrectly. Anything outside a normal distribution and they don't make sense.
All this is irrelevant for Rossi, his use of the term is informal and non-scientific. (Also, with high probability, inaccurate).
-
As always then things like 5 sigma can be applied correctly or incorrectly. Anything outside a normal distribution and they don't make sense.
Are you describing a non-parametric approach, above, which is then used to determine whether to assume a Gaussian distribution? -
I don't know anymore if this forum pretends paid shills and bots don't exist for whatever reason, or if people administrating it are completely dense
Anyway, I'm starting to understand why the Alexander Grothendiecks quit the scientific world never to come back, they've seen devils up close and it doesn't suit them
-
Six Sigma is a common term and concept in product manufacturing, championed by Motorola:
https://en.m.wikipedia.org/wiki/Six_Sigma
Hence Rossi's "5 Sigma" may simply refer to a less strict variant of this concept.
-
There is a lot of speculation of what Andrea Rossi is doing to get his 5 sigma. I am quite sure it has to do with (in no particular sequence):
- reliability
- performance
- safety
- producibility
- controllability
I am sure there are more. So how would AR test all these different aspects in a relatively short period. How would you measure each of these on 5 Sigma?
I have no idea, I have never had to do with 5S, but maybe someone is more familiar with this?
He has mentioned more often that he had 3 QuarkX's running, but that may have changed in time... -
The known structure of the problem may make a simple normal distribution incorrect. That could be apparent from the data alone (say if bimodal) and I would also not rule out a hypothesis from which an unusual distribution could be derived. In this case we could separate "probability any given reactor works" and "probability that LENR is real given the totality of data". The two need not be the same.
-
Damn Right Man,
I think the basic operation of the hot cat and low temp E-Cats are pretty simple -- although usually requiring more than a one off test by a replicator. From Rossi's point of view, IH has received his IP and could be doing anything they want with the information. I personally think it is likely (there is no way to be sure) IH has indeed replicated the low temp E-Cat and high temp E-Cat in a laboratory somewhere. And I don't think it would take an excessive amount of labor or money for anyone else to do the same. I think Rossi realized this and produced the Quark as having an insurance policy to make sure he had SOME IP that wasn't already publicly known.
If IH had paid Rossi for the year long test (or whatever it actually was), I think he would have had no choice legally except to share the Quark IP with them. He is using their non-payment (which may or may not have been appropriate) as a justification for witholding the IP from them. This probably infuriates them, because it represents a method of directly producing electricity with very high COP without the need for additional conversion equipment (steam generators, thermoelectric converters, etc). The original "hot cat" was capable of changing the world, in my opinion, but I think the Quark represents a leap beyond the production of heat. Basically, it would be like a fuse that produced more power out than you put in -- in addition to producing a high level of heat.
-
Of course the various ecats are claimed to make no high energy radiation -- they are nothing but electrical heaters. Except for the QuarkX which seems to be purely fiction. What Moletrappers have dubbed Rossifiction. Rossi is not "withholding" IP. He has never had any IP and that includes not only the ecat but goes all the way back to the Petroldragon and thermoelectric scams. No IP. Only scam.
-
Of course the various ecats are claimed to make no high energy radiation -- they are nothing but electrical heaters. Except for the QuarkX which seems to be purely fiction. What Moletrappers have dubbed Rossifiction. Rossi is not "withholding" IP. He has never had any IP and that includes not only the ecat but goes all the way back to the Petroldragon and thermoelectric scams. No IP. Only scam.
MaryYugo, Rossi is headed for a tragedy of epic proportions, a raft of pain that will make your heart spill over with joy. But all your arguments will not produce this plague that will fall onto Rossi's head. Rossi will do it to himself, it just a matter of time; so be patient with beautiful patience. Having beautiful patience does not mean you cannot feel sad or cry; beautiful patience is knowing that the calamity that will befall Rossi will be part of fates great plan; a plan that you had no hand in bring to pass. So be happy, sit back, and enjoy all the fun.
-
Since I didn't know what 5 sigma means in this context, I googled around a bit and found this article:
https://blogs.scientificameric…ons/five-sigmawhats-that/
QuoteIn short, five-sigma corresponds to a p-value, or probability, of 3x10-7, or about 1 in 3.5 million. This is not the probability that the Higgs boson does or doesn't exist; rather, it is the probability that if the particle does not exist, the data that CERN scientists collected in Geneva, Switzerland, would be at least as extreme as what they observed. "The reason that it's so annoying is that people want to hear declarative statements, like 'The probability that there's a Higgs is 99.9 percent,' but the real statement has an 'if' in there. There's a conditional. There's no way to remove the conditional," says Kyle Cranmer, a physicist at New York University and member of the ATLAS team, one of the two groups that announced the new particle results in Geneva on July 4.
So if we apply this to the E-cat, we are talking about a 1 in 3.5 million chance that the Rossi effect is not real given the result obtained with the QuarkX. Not a 1 in 3.5 million chance that the QuarkX works.
