by lkcl » Tue Jan 31, 2017 8:20 am
http://physics.nist.gov/cgi-bin/cuu/Value?bghi fredi (and joy),
i was very excited to read your paper, particularly having read andrew worsley's extremely comprehensive mass-analysis work. i am very familiar with the nist.gov list of physical constants, particularly their relative standard uncertainty, because in any "numerical-only" statistical inference work, the relative uncertainty is *directly* the level of accuracy to which explorations may be made.
for example: in working with andrew to narrow down some of the studies we were doing to below 1e-7 accuracy, there was, quite obviously, absolutely no point in using *anything* which contained a fundamental constant that had a relative uncertainty of 1e-8 (i actually set the cutoff at 2 orders of magnitude lower than that, to 1e-10, so that i could reasonably argue a statistical correlation discrepancy of 1 in a 1000, and i wasn't really very happy about that low a correlation... but you get the point i am sure).
all of that is merely background to put into context the following question:
i assume that G (as used in the paper) is the "Gravitational constant" (if i am wrong about that please ignore this question entirely!) how can you justify the conclusions that you're making (particularly the calculations in the Appendix) which are to such extreme levels of accuracy (requiring arbitrary floating-point precision), and *particularly* when you appear to be subtracting two very large numbers to find a very very small one (a problem which any computer scientist will tell you is asking for trouble, but you have dealt with it very well by going to arbitrary-precision floating-point... *except*....) that the Gravitational Constant's relative standard uncertainty is (according to the link above) 4.7 x 10^-5 and i don't see _any_ indication in the paper that that's taken into account.
now, i may just be ignorant of many of the conventions under which papers are written, such that there are assumptions which i am simply not aware of, but my computer science background and training is making me jump up and down here with a huuuuge red flag
looking at these "rules"
http://web.uvic.ca/~jalexndr/192UncertRules.pdf i note that even the square/cube (etc.) stuff, you can only get away with dividing the relative uncertainty by two (or three, or whatever the power is).
in essence: is there something fundamental that i missed, here? have i misinterpreted something which makes the question moot?
http://physics.nist.gov/cgi-bin/cuu/Value?bg
hi fredi (and joy),
i was very excited to read your paper, particularly having read andrew worsley's extremely comprehensive mass-analysis work. i am very familiar with the nist.gov list of physical constants, particularly their relative standard uncertainty, because in any "numerical-only" statistical inference work, the relative uncertainty is *directly* the level of accuracy to which explorations may be made.
for example: in working with andrew to narrow down some of the studies we were doing to below 1e-7 accuracy, there was, quite obviously, absolutely no point in using *anything* which contained a fundamental constant that had a relative uncertainty of 1e-8 (i actually set the cutoff at 2 orders of magnitude lower than that, to 1e-10, so that i could reasonably argue a statistical correlation discrepancy of 1 in a 1000, and i wasn't really very happy about that low a correlation... but you get the point i am sure).
all of that is merely background to put into context the following question:
i assume that G (as used in the paper) is the "Gravitational constant" (if i am wrong about that please ignore this question entirely!) how can you justify the conclusions that you're making (particularly the calculations in the Appendix) which are to such extreme levels of accuracy (requiring arbitrary floating-point precision), and *particularly* when you appear to be subtracting two very large numbers to find a very very small one (a problem which any computer scientist will tell you is asking for trouble, but you have dealt with it very well by going to arbitrary-precision floating-point... *except*....) that the Gravitational Constant's relative standard uncertainty is (according to the link above) 4.7 x 10^-5 and i don't see _any_ indication in the paper that that's taken into account.
now, i may just be ignorant of many of the conventions under which papers are written, such that there are assumptions which i am simply not aware of, but my computer science background and training is making me jump up and down here with a huuuuge red flag :)
looking at these "rules" http://web.uvic.ca/~jalexndr/192UncertRules.pdf i note that even the square/cube (etc.) stuff, you can only get away with dividing the relative uncertainty by two (or three, or whatever the power is).
in essence: is there something fundamental that i missed, here? have i misinterpreted something which makes the question moot?