2 ms·

I see the paper says: >"[MOND predicts] the expected velocity dispersion of NGC1052–DF2 is ... 20 km s−1, a factor of two higher than the 90% upper limit on th

by nonbel 9y ago

I see the paper says: >"[MOND predicts] the expected velocity dispersion of NGC1052–DF2 is ... 20 km s−1, a factor of two higher than the 90% upper limit on the observed dispersion." http://imgsrc.hubblesite.org/hvi/uploads/science_paper/file_attachment/317/van_Dokuum_paper_submitted_to_Nature.pdf http://imgsrc.hubblesite.org/hvi/uploads/science_paper/file_... So they chose to use a 90% CI along with assuming the error is normally distributed, etc. Looking at figure 3b it seems that if they instead arbitrarily chose to use a 95% CI that 20 km/s would be within the interval. Ie, these results appear to be moderately unlikely if MOND were correct but extremely unlikely if "dark matter" is correct. Thus, according to Bayes' rule, the probability that MOND is correct has increased.

nonbel 9y ago
Not sure where the downvotes are coming from. The reasoning that this evidence is moderately unlikely given that MOND is true can be found in the parent post. The authors worked it out to ~1/10 to 1/20. They don't put a number on how unlikely it would be given dark matter is true, but it sounds pretty extreme: >Merritt remarks: "There is no theory that predicts these types of galaxies — how you actually go about forming one of these things is completely unknown." http://www.spacetelescope.org/news/heic1806/ http://www.spacetelescope.org/news/heic1806/ >"And NGC 1052-DF2 appears to have none. Or at best, about as much dark matter as normal matter. And that just doesn't make sense." http://www.syfy.com/syfywire/what-is-this-galaxy-doing-without-a-dark-matter-halo http://www.syfy.com/syfywire/what-is-this-galaxy-doing-witho... The denominator of Bayes rule for Pr(MOND|data) is just the sum of all Pr(Theory) x Pr(data|Theory) where the main theory is "dark matter" and second is MOND (everything else is some small value epsilon), while the numerator is Pr(MOND) x Pr(data|MOND). Where is the mistake?

by nonbel 9y ago

nonbel 9y ago

Not sure where the downvotes are coming from. The reasoning that this evidence is moderately unlikely given that MOND is true can be found in the parent post. The authors worked it out to ~1/10 to 1/20. They don't put a number on how unlikely it would be given dark matter is true, but it sounds pretty extreme: >Merritt remarks: "There is no theory that predicts these types of galaxies — how you actually go about forming one of these things is completely unknown." http://www.spacetelescope.org/news/heic1806/ http://www.spacetelescope.org/news/heic1806/ >"And NGC 1052-DF2 appears to have none. Or at best, about as much dark matter as normal matter. And that just doesn't make sense." http://www.syfy.com/syfywire/what-is-this-galaxy-doing-without-a-dark-matter-halo http://www.syfy.com/syfywire/what-is-this-galaxy-doing-witho... The denominator of Bayes rule for Pr(MOND|data) is just the sum of all Pr(Theory) x Pr(data|Theory) where the main theory is "dark matter" and second is MOND (everything else is some small value epsilon), while the numerator is Pr(MOND) x Pr(data|MOND). Where is the mistake?