Potentially major ALMA array discovery? Options

[mcaplinger]

Tough crowd. And I must have missed the memo that said peer review could be done over Twitter.

The system will sort this out. I don't think you want to introduce a chilling effect where interesting results, even if premature, can't be put out on arxiv. The authors are taking chances of being criticized, but those are their chances to take.

[Deimos]

It is true that many implausible things happen; not so much that anything is possible, but that extremely implausible things remain possible. But that doesn't mean we have to accept all claims with equal credulity.

Suppose that the kind of thing the paper claims to discover is a real kind of thing. Let's suppose there is exactly one such thing. There are a few chances in a billion that a specific ALMA aim contains said object. Maybe a few chances in a million that a similarly sensitive ALMA observation contains it (yes, that's purely speculative, but generous, I think). On the other hand, assuming the errors in the ALMA data are exactly Gaussian[*], there's about one chance in 13 of a >4 sigma false positive per observation[**]. So, even if a thing like they propose really exists, the chances are >>1,000,000 to 1 that a specific 4-sigma detection is a false positive.

OK, suppose there are 1000 such things, all staying just under WISE detectability (in itself, somewhat contrived and unlikely). Even if that is true, there is still a >>1000:1 preponderance of false positives.

There's a reason for peer review. And there's a reason you don't claim a discovery just because you can fit a line to two points. You need to try that third point before you've tested a hypothesis. Otherwise you reduce yourself to arguing about your idea of implausible vs. mine, or non-Gaussian statistics, or whether the instrumental effects are handled right in the 99.99th percentile cases, or coincidence of different objects, etc.

Of course, if we crank our supposition up to hundreds of thousands of real such objects, the report borders on plausible; but Mike Brown says we are all dead then, so we likely shouldn't be arguing over such worldly matters.

[*] If the instrument were a megapixel CCD detector of visible light, a SNR>20:1 false positive would not be a big surprise, given terrestrial radiation, cosmic rays and such. Many measurements have distinctly non-guassian tails to their errors.

[**] They indicate about 50x50 resolution elements, or 2,500 measurements, per FOV in the paper; a >4-sigma positive outlier is expected around 1 per 32,000 measurements.

[Herobrine]

There are quotes floating around from Mike Brown suggesting this is indeed the case, but I haven't seen any actual numbers.

If you're referring to the several tweets Brown made over a span of a half hour, I don't think they were comments that came from any serious work or thought; he seemed to be just tweeting his immediate reaction, crunching a few rough numbers so he could include a figure in his tweets. The tweets are quoted on the first page of this thread. Checking now, he responded on Twitter to a question about where the 200,000 figure came from, with

it was based on a wild but generous guess of how much sky ALMA's covered to that depth, twice.

https://twitter.com/plutokiller/status/676182202368045057
His response to the person who asked how it would kill everyone was the tweet:

crushed by planets.

https://twitter.com/plutokiller/status/675472733271035905

I don't share the tendency of many to use how 'lucky' something would have to have been as an argument against it, but I'd think that recognition of that should be enough to cause one to look back over one's findings and verify that the data strongly support a claim before making it, particularly when the claim is as absolute as "ALMA discovers the most distant object of the solar system".

[JRehling]

I don't share the tendency of many to use how 'lucky' something would have to have been as an argument against it

If you find a footprint that was made in a random London basement over 100 years ago, it might be Jack the Ripper's footprint, but the extreme improbability of that is a reason not to say that the footprint solves the Jack the Ripper case.

It's perfectly fine to say that it's a footprint and start to figure out whose it might have been. And that just might end up solving the Jack the Ripper case, but that's a claim to make much, much later in the process, not at the start.

[katodomo]

They do address WISE and question why it would only be detectable now, suggesting that there was too much glare from AC or that it may have been too cold for WISE to catch.

If it's detectable now (and of sufficient magnitude) it should be hiding somewhere in Gaia's heaps of current data. Probably in the part scheduled for processed release in 2022...

[Herobrine]

If you find a footprint that was made in a random London basement over 100 years ago, it might be Jack the Ripper's footprint, but the extreme improbability of that is a reason not to say that the footprint solves the Jack the Ripper case.

It's perfectly fine to say that it's a footprint and start to figure out whose it might have been. And that just might end up solving the Jack the Ripper case, but that's a claim to make much, much later in the process, not at the start.

I don't know how that's supposed to relate to what I wrote.

I said I don't use how 'lucky' something would have to have been as an argument against it. That doesn't mean I think every random, baseless proposition is true. In your Jack the Ripper footprint scenario, you have no reason to think it's his in the first place; that's the argument against saying it's his. How 'lucky' you'd have to be doesn't factor into it. Amend the scenario to have countless pieces of evidence in the basement, all strongly suggesting the footprint belongs to Jack the Ripper, and you'd probably rethink the notion that the 'luck' you would have to have to stumble upon Jack the Ripper's footprint is a reason not to say that the footprint is his.

The same thing would happen to your scenario if, instead of additional evidence, there were only two people who ever lived in London, one of the two was Jack the Ripper, and London itself consisted only of 2 basements. You wouldn't have to be very 'lucky' at all in that situation; in fact, you have a ~50% chance of any given footprint being his. That's still not a valid argument for or against it actually being his. Even in this scenario, you don't have any reason to claim that it's his. The requisite 'luck' changed drastically between the two scenarios, but the basis for such a claim is exactly the same as the original scenario; there isn't one.

If requisite 'luck' is an argument against things, then we should all be very skeptical of lottery winners.

[fredk]

We can talk about two possibilities for the alpha Cen observation:

1.) They caught a distant solar system body in two positions. The likelihood can be estimated as I wrote before given the estimated number density of bodies. There are indications from Brown that this is extremely unlikely (though we haven't seen all the numbers).

2.) They caught a transient on some distant stellar or galactic source, or some instrumental glitch, for the high S/N detection, plus a noise fluctuation for the low S/N observation. The odds for this are unclear, but likely not negligible, especially for a new instrument.

If actual quantitative estimates give a tiny probability for 1.) but order 50%, say, for 2.), then that tells you clearly which possibility is most likely. This is how we can say how "lucky" something is can be an argument against it. The key is that you have an alternative that is likely. This is done all the time in science: there often are alternative explanations to observations that can be discounted as extremely unlikely relative to other explanations. I see an elephant unlike any seen before in the wild. Far more likely that it's a one-off mutant than a member of a previously unseen new species.

[Deimos]

A Bayesian outlook is quite compatible with lottery winners--if there is sufficiently good evidence they won the lottery. But if someone asks for a check but only proves they match the first 2 numbers, you should be skeptical. A SNR~4 observation still leaves false positive as the most likely explanation. A very high confidence observation still leaves coincidence of different objects as a more likely explanation, when fitting a line to only 2 points. Sure, reality will not bend to conform to the most likely explanation; but the most likely explanation will be right a lot more often than highly unlikely ones. So 'lucky' shouldn't be used to cross something off the list of competing hypotheses. But it very well should be considered, and balanced with how 'lucky' you'd have to be to get the observation from other hypotheses, if you're in the business of making predictions (i.e., science) or announcing shocking discoveries. Another couple observations in the band capable of SNR>20, and the discussion of how 'lucky' they were would fall by the wayside. But it is very, very relevant now, because 'noise' wouldn't have to be nearly so lucky.

[Mongo]

Tentative planetary orbital constraints of some scenarios for the possible new Solar System object recently discovered with ALMA

Some of the scenarios envisaged for the possible new Solar System object, whose discovery with the ALMA facility has been recently claimed in the literature, are preliminarily put to the test by means of the orbital motions of some planets of the Solar System. It turns out that the current ranges of admissible values for any anomalous secular precession of the perihelion of Saturn, determined in the recent past with either the EPM2011 and the INPOP10a planetary ephemerides without modeling the action of such a potential new member of the Solar System, do not rule out the existence of a putative Neptune-like pointlike perturber at about 2500 au. Instead, both a super-Earth at some hundreds of au and a Jovian-type planet up to 4000 au are strongly disfavored. An Earth-sized body at 100 au would have a density as little as ∼0.1−0.01 g cm−3, while an unusually large Centaur or (Extreme) Trans Neptunian Object with linear size of 220−880 km at 12−25 au would have density much larger than ∼1 g cm−3.

[Spock1108]

I think this explanation is one of the most consistent ...

[alan]

Abstract of the Alpha Centauri paper now reads:

We recently announced the detection of an unknown submillimeter source in our ALMA observations of alpha Cen AB. The source was detected in two epochs, a strong detection at 445~GHz and one at lower significance at 343.5~GHz. After valuable feedback of the community, it turns out that the detection at 343.5~GHz could not be reproduced with a different reduction software nor with fitting within the (u,v)-plane. The detection at 445~GHz has been further confirmed with modeling of the (u,v)-data and was shown to be robust at >12σ, confirming our detection of this unknown source. However, based on only one epoch, further analysis and preferably new data are needed, before publication of an article in which the nature of the new source can be discussed. The analysis has indicated the importance of both (u,v)-plane fitting and alternative data reduction when dealing with low signal to noise source detections.

Under comments: withdrawn until further data is available