I wonder if the resolution is that the first is supposed to read ‘the ratio of Sun-like stars to Earth-sized planets orbiting such a star within its habitable zone is around 5:1’. That would double-count stars with two such planets.
“Analyses of the Kepler results shows that 7-15% of the Sun-like stars have an Earth-sized planet within their habitable zone [Petigura et al., 2014]”
Perhaps,
“Analyses of the Kepler results yields a minimum fraction of Sun-like stars with an Earth-sized planet within their habitable zone, of 11 +/- 4% [Petigura et al., 2014]”
Notice how much more labored and pedantic your version is—the sort of writing that one would not do unless one could see into the future that there would be nerds somewhere nitpicking exactly that sentence.
It is more labored, because it’s attempting to convey a more complicated concept. However, the distinction is not pedantic. This is saying ‘there is one fence near here, somewhere within this range’. The other statement means ‘there are two fences here enclosing this range.’. These are not at all interchangeable statements.
I dunno. Not an astronomer. But there are lots of different strategies for measuring things, which come with their own particular strengths and weaknesses, so I wouldn’t be surprised if some available measures of some fraction had different inherent bounds or precisions based on available data.
(For example, in genomics, it’s not uncommon to have a lower bound with a confidence interval; in fact, every GCTA study using SNPs produces a lower bound with a somewhat loose confidence interval, and this has tripped up some commentators who, upon observing an estimated heritability of, say, 0.25-0.30 for intelligence from one study, triumphantly declare that the glass is more than half-empty—forgetting that it’s a lower bound, and different GCTAs using differing levels of comprehensiveness of SNPs will turn in different lower bounds and so one could easily have a GCTA estimate 0-0.20 and another 0.25-0.30, in contradistinction to twin studies with heritability of 0.5 or higher—based on how many SNPs were included and how many samples there were!
Or to take a physics example from my reading yesterday, Meehl 1990. Meehl, discussing philosophy of science & statistics, notes that in the book Atoms (early 1900s) are covered 13 different ways of estimating Avogadro’s number which result in different numbers of the same magnitude but that treated in terms of random sampling error, the 13 ways would yield confidence intervals that would often exclude each other’s. Surely, he asks, we would not reject the 13 consilient arguments for the existence of atoms solely because of this slight discrepancy, and instead regard the slight disagreement as purely springing from systematic error such as the differing approximations and simplifying assumptions made?)
‘7-15%’ does not look like a lower bound to me.
I wonder if the resolution is that the first is supposed to read ‘the ratio of Sun-like stars to Earth-sized planets orbiting such a star within its habitable zone is around 5:1’. That would double-count stars with two such planets.
If there is uncertainty in the lower bound produced by an argument, how else would you write it?
Certainly not
“Analyses of the Kepler results shows that 7-15% of the Sun-like stars have an Earth-sized planet within their habitable zone [Petigura et al., 2014]”
Perhaps,
“Analyses of the Kepler results yields a minimum fraction of Sun-like stars with an Earth-sized planet within their habitable zone, of 11 +/- 4% [Petigura et al., 2014]”
Notice how much more labored and pedantic your version is—the sort of writing that one would not do unless one could see into the future that there would be nerds somewhere nitpicking exactly that sentence.
It is more labored, because it’s attempting to convey a more complicated concept. However, the distinction is not pedantic. This is saying ‘there is one fence near here, somewhere within this range’. The other statement means ‘there are two fences here enclosing this range.’. These are not at all interchangeable statements.
In what setup would the difference between the two be measurable?
I dunno. Not an astronomer. But there are lots of different strategies for measuring things, which come with their own particular strengths and weaknesses, so I wouldn’t be surprised if some available measures of some fraction had different inherent bounds or precisions based on available data.
(For example, in genomics, it’s not uncommon to have a lower bound with a confidence interval; in fact, every GCTA study using SNPs produces a lower bound with a somewhat loose confidence interval, and this has tripped up some commentators who, upon observing an estimated heritability of, say, 0.25-0.30 for intelligence from one study, triumphantly declare that the glass is more than half-empty—forgetting that it’s a lower bound, and different GCTAs using differing levels of comprehensiveness of SNPs will turn in different lower bounds and so one could easily have a GCTA estimate 0-0.20 and another 0.25-0.30, in contradistinction to twin studies with heritability of 0.5 or higher—based on how many SNPs were included and how many samples there were!
Or to take a physics example from my reading yesterday, Meehl 1990. Meehl, discussing philosophy of science & statistics, notes that in the book Atoms (early 1900s) are covered 13 different ways of estimating Avogadro’s number which result in different numbers of the same magnitude but that treated in terms of random sampling error, the 13 ways would yield confidence intervals that would often exclude each other’s. Surely, he asks, we would not reject the 13 consilient arguments for the existence of atoms solely because of this slight discrepancy, and instead regard the slight disagreement as purely springing from systematic error such as the differing approximations and simplifying assumptions made?)