Arp objects, QSOs, Statistics

In this post I'll continue to look at the BAC calculation on its own, 'blind' to where it came from, what interpretations can (or cannot) be drawn from the 'probabilities', etc.

I have already explored how the Karlsson peaks behave.

How do the other three peaks (Amaik, regular, and DRD) behave?

Recall that the Amaik peaks are: 0.04, 0.43, 0.78, 1.13, 1.56, 2.33, and 2.43.

And the regular peaks are: 0.38, 0.75, 1.13, 1.50, 1.88, 2.25, and 2.63.

And the DRDS peaks {drum roll please ...*}: 0.22, 0.63, 1.03, 1.36, 1.63, 1.88, 2.58.

Specifically, under what conditions can they give a 'BAC probability' of zero?

Is it possible for any of them to produce 'BAC probabilities' > 1?

We could play with some numbers, explore how the 'BAC probabilities' change with varying inputs etc ....

... or we could extend the results we already have, from our investigation of the 'Karlsson peaks'.

So, it is possible for a 'BAC probability' to be zero for any of these three other kinds of peaks, and under the same circumstances (i.e. an input z is the same as the value of a peak).

Anyone not see why this follows, logically?

Whether 'BAC probabilities' can be >1, for any of the three other kinds of peaks may not be so straight-forward, but perhaps we can say, with a reasonable degree of confidence, that this can happen for the Amaik peaks, but not the other two (based on what we learned from the investigation of the Karlsson peaks).

Anyone care to explain why?

Comments?

Next: for each of these three other kinds of peaks, do the 'BAC probabilities' also get smaller as the number of input 'redshifts' increases?

* I think this is the first time I've presented their values!
 
In this post I'll continue to look at the BAC calculation on its own, 'blind' to where it came from, what interpretations can (or cannot) be drawn from the 'probabilities', etc.

Using the same 300 random numbers in [0, 3] as I used to explore the behaviour of the 'Karlsson peaks' with the 'BAC probabilities', and the same 295, 291, and 281 sets of 6, 10, and 20 dzs (respectively) ...

I found the nearest 'Amaik peak'/regular peak/'DRDS peak' to each random number, and calculated the 'dz' for each.

There are 10/10/4 dz's < 0.005 (Amaik, regular, DRDS, respectively).

The 'BAC probabilities' for 'Amaik peaks' range from 7.1x10-8 to 1.65, with an average of 0.074 (sets of six, two significant figures); 4.8x10-9 to 0.82 (ten), with an average of 0.011; and 3.3x10-13 to 0.00041, with an average of 8.7x10-6 (20).

For 'regular peaks', The 'BAC probabilities' range from 1.6x10-6 to 0.47, with an average of 0.022 (sets of six, two significant figures); 1.8x10-8 to 0.065 (ten), with an average of 0.0013; and 1.7x10-13 to 0.00032, with an average of 3.3x10-6 (20).

For 'DRDS peaks', The 'BAC probabilities' range from 2.0x10-6 to 0.36, with an average of 0.023 (sets of six, two significant figures); 3.6x10-8 to 0.14 (ten), with an average of 0.0021; and 1.3x10-12 to 0.00012, with an average of 1.7x10-6 (20).

So it seems the more input 'redshifts' to the 'BAC calculation', the lower the 'BAC probability', just as with the 'Karlsson peaks'.
 
Time to start looking at the 'BAC probabilities', and 'BAC calculations', and 'BAC approach' in a broader way.

One thing is immediately obvious: 'BAC probabilities' are not probabilities, in the normal meaning of that word ... because valid, legitimate inputs can produce a 'probability' that is > 1.

Further, because the 'BAC probabilities' are calculated, in the final stage, by multiplying an always positive factor by many 'dzs', a 'zero probability' follows if any one of the dzs is zero. And if these dzs are given to only two decimal places, such a condition will arise in ~1% of (nearly) all cases (to a factor of a few)*, assuming the input z's are distributed more or less randomly in [0, 3].

Too, there's nothing special about the 'Karlsson peaks' ... many different kinds of 'peaks' will produce similar results.

So looking back at my various test sets (group a), group b), sets ONE through SEVEN; the several 'Arpian' cases), it should come as no surprise that you can't tell which are from quasars predominantly along the minor axes of certain bright spirals (including the 'Arpian faves'), which are predominantly along the major axes of the same spirals, which are along random directions in random fields, and which are from random numbers generated on my PC.

Clearly, the calculation of 'BAC probabilities' does not say anything about the likelihood of quasars being preferentially near Kalsson peaks and predominantly along the minor axes of certain bright spirals!

Of course, I may have simply misunderstood the steps to be used to calculate the 'BAC probabilities', or I may have not interpreted the results of such calculations correctly, or BAC may have made a mistake in deriving the formula, or .... If so, I trust that BeAChooser will set the record straight, forthwith.

Comments?

* it's easy enough to create pathological cases where this doesn't happen
 
I have to agree with you. There is something wrong when a probability calculation can give values > 1.
As for using Karlsson peaks, not only will other sets of peaks produce similar values (e.g. those from Bell’s decreasing intrinsic redshift (DIR) model) there is also doubts about the existence of Karlsson peaks. Have a look at Critical Examinations of QSO Redshift Periodicities and Associations with Galaxies in Sloan Digital Sky Survey Data. This uses subsets of the SDSS data. There is a later paper using all of the SDSS data by Bell that supports his peaks (IMHO more suggestive of the properties of the SDSS sample than an actual periodicity).
 
Reality Check said:
I have to agree with you. There is something wrong when a probability calculation can give values > 1.
As for using Karlsson peaks, not only will other sets of peaks produce similar values (e.g. those from Bell’s decreasing intrinsic redshift (DIR) model) there is also doubts about the existence of Karlsson peaks. Have a look at Critical Examinations of QSO Redshift Periodicities and Associations with Galaxies in Sloan Digital Sky Survey Data. This uses subsets of the SDSS data. There is a later paper using all of the SDSS data by Bell that supports his peaks (IMHO more suggestive of the properties of the SDSS sample than an actual periodicity).
Click to expand...
Thanks for this! :)

I have seen a case made that several classes of 'quasars are ejected from AGNs' model are not knocked out by this Tang and Zhang paper, and such models may not be knocked out by some generalised version of their method. I don't recall all the details, but it goes something like this:

* quasars are heterogeneous; some are at cosmological distances consistent with their (Hubble) redshifts, others are ejected by (local) galaxies

* the ejector galaxies of the 'ejectee quasars' may no longer be active, so all spirals need to be examined (not just Seyferts)

* 'ejectee quasars' may be found quite a long way (on the sky) from their 'parents', up to several tens of degrees

* there is no quantisation of ejection/'intrinsic' redshift.

Not surprisingly, these kinds of models are almost impossible to falsify! Particularly as they have no physics behind them (they are derived entirely from 'empirical observations').

However, there is one sure test .... quasar proper motions (and parallaxes); in all these models the ejectee quasars are moving at very high speeds, relative to their parents (and so to us); those from the nearest parent (M81, say) will have, on average, proper motions that GAIA should easily detect (assuming that changes in the location of the photocentre, due to new jet blobs, for example, can be controlled). Stay tuned!
 
Reality Check said:
I have to agree with you. There is something wrong when a probability calculation can give values > 1.
As for using Karlsson peaks, not only will other sets of peaks produce similar values (e.g. those from Bell’s decreasing intrinsic redshift (DIR) model) there is also doubts about the existence of Karlsson peaks. Have a look at Critical Examinations of QSO Redshift Periodicities and Associations with Galaxies in Sloan Digital Sky Survey Data. This uses subsets of the SDSS data. There is a later paper using all of the SDSS data by Bell that supports his peaks (IMHO more suggestive of the properties of the SDSS sample than an actual periodicity).
Click to expand...

If I understand correctly, couldn't this paper be Arp's control case? At first glance, they have just looked at QSO's and galaxies, both at random and with some preferences towards AGN's and such.

There are other limitations which they note in their analysis, but doesn't this seem to be the control case?

Doesn't this help Arp, and not hinder?
 
It depends on the samples, they say a certain number of sets. But they don't give much detail. especialy regards the standard deviation of association crossed by arc seconds from the galaxy.

It doesn't make reference if the Arp sample also meets the criteria for 'random sample pairs'.

So I would say no, and to do so would be meta-analysis which is a pain.

Once the survey is close to complete, random samples could be taken to generate the arc second/association with galaxies figure, which would then be the base line for Arpists to compare to. Sample should be at least 1000 galaxies and 1000 random spots.

This is not that.

the paper appears to mainly be about z periodicity and Bell's hypothesis.
 
Last edited:
Wangler said:
If I understand correctly, couldn't this paper be Arp's control case? At first glance, they have just looked at QSO's and galaxies, both at random and with some preferences towards AGN's and such.

There are other limitations which they note in their analysis, but doesn't this seem to be the control case?

Doesn't this help Arp, and not hinder?
Click to expand...
The paper was published in 2005 and Arp' example cases are before then. In fact I cannot find any since 2005. Nor can I find any Arp paper citing this paper.
The paper basically throws doubt on the existence of Karlsson peaks. These peaks are used in Arp's papers. If they do not exist then his association of quasars with bright galaxies is in doubt.
 
Quick blast from the past.

Found this short six page paper, which addresses these issues in terms of Bayes theorem. Someone more versed in statistics/probabilities than I should have a look at it and say what they think.

Statistics – Tool or Weapon?
Statistics – Tool or Weapon?

The application and misapplication of Bayes’ theorem

There is an old saying: “Statistics don‟t lie… but statisticians sometimes do.” This maxim is a corollary of the well known fact that „Garbage In produces Garbage Out‟ (GIGO). It reaffirms the observation that, no matter how sophisticated any given computer program may be, if you feed it input data that is flawed, the output you will get will also be flawed. Despite this however, most people are usually impressed by complication: the more complicated the computerized algorithm, the more persuasive the results are for anyone who is gullible. The effectiveness of complicated statistics is a case in point.

Most of us have heard the terms a priori and a posteriori and know they refer respectively to things that precede or follow an event. But most people do not know how to calculate nor to interpret an „a posteriori probability‟ even though they may have occasionally heard that term. Such quantities are often used in science and we ought to know how to interpret them and how to judge the appropriateness of using them. For this we can turn to Bayes‟ theorem. [....]

First of all, it wasn‟t an a posteriori probability. As we have seen, an a posteriori probability quantifies how much you can trust the results presented by some kind of test. Arp was not performing any kind of test. He was simply comparing the observed density of a certain type object in one place on the sky contrasted with any other location. Arp „saw the tiger in the tent with his own eyes‟ – he did not need to perform any kind of test in order to verify those observations.

Throwing around the descriptor „a posteriori‟ in a pejorative attempt to belittle Arp‟s work, clearly demonstrates that the critic either does not understand probability theory – or hopes that we don‟t.

But we do.
Click to expand...


Sounds good to me :)

Anything wrong with it?
 
It has nothing to do with this thread since Arp does use not Bayesian statistics.

He does not really do statistics. He merely gives examples and calculates probabilities.

One point:
"For example, astronomer Halton Arp has presented a long series of images of unusual concentrations of BL Lac objects relatively near Seyfert galaxies."
I would not call 7 or 8 examples a "long series".

IMHO: This issue will hang around until Arp (or another astronomer with some spare time) does the sensible test.
Repeat the analysis for unusual concentrations of BL Lac objects relatively near non-Seyfert galaxies, e.g. dwarf galaxies. If a similar number of examples can be found then we have that quasars can be emitted from ordinary galaxies which contradicts Arp's hypothesis.
​
The "a posteriori" part of Arp's statistics is that he starts with the hypothesis that quasars are emitted from Seyfert galaxies and then looks for examples that match his hypothesis.

P.S. Linking stuff from Don Scott's crackpot Electric Sky site is not helping your repuation! :D
 
To more clearly see the problem with Arp's methodology consider the following adaption of it:
Look for images of unusual concentrations of quasars relatively near an arbitary empty point such as the center of the concentration (making sure that there is no galaxy that point).
If there are such images then run Arp's probability calculation. You will get the same magntiude of probability since the calculation does not depend on the nature of the central object.
​
Would Arp then conculde that quasars are emitted from empty space?

P.S. Maybe the linked article is even further off topic. This thread is about quasars. Don Scott mentions BL Lac objects which are a type of active galaxy.
 
DeiRenDopa said:
robinson said:
7 post in a row. And nobody responded. I know you are new, so let me give you a clue. If nobody is responding, let it die a quiet and dignified death.
Click to expand...
What, and miss out on another page or five of posts I can add to the 'Questions BAC has run away from' thread (or whatever it's called)?

You of all JREF forum members should surely appreciate the value of holding people to account for what they write, shouldn't you*? I mean, especially given the venom, bombast, ridicule, and more heaped upon those who didn't go along with BAC's certainty over what we might call, in shorthand, 'the Arp Karlsson peak quasars'?

And surely if there's anything to 'Arp's statistics', given how dramatically it would change not only astronomy but also physics, don't you think it worthwhile to run it to ground (and show just how full of errors of many different kinds this idea/these ideas is/are ... or not)?

* at least in this section of the forum, and on matters with clear scientific content.
Click to expand...
I re-read this long thread in response, ultimately, to a PM I recently got (thanks sender, you know who you are).

I'm quoting this post, and bumping this whole thread, to make a very important point.

BAC (BeAChooser) did something that I don't think any of the many 'mainstream astrophysics/cosmology/astronomy/space science is rong!' JREF members who have posted so avidly these past few years have done ... not Zeuzzzzzz, not MM, not brantc, not Anaconda, not sol88, not solrey*, not robinson, not Bjarne, and most recently (perhaps) not Siggy_G (though the night is still young for him).

And what amazing thing did BAC do, to make him stand out among this illustrious crowd?

Why he actually proposed something testable, as in quantitatively testable, in an objective, independently verifiable way! :jaw-dropp

That he spent dozens of long posts obfuscating, blustering, and generally engaging in diversionary tactics; that it turns out his test failed, spectacularly; that he never returned (and so never admitted the spectacular failure); that the particular topic is rather narrow; and so on - none of this matters.

What matter is that he tried to do some actual science. :cool: :)

Why is it that what BAC did is so rare, among JREF members with these inklings (at least in this section of JREF)?

Why do his 'alternatives' peers seem so much happier spending endless hours writing posts full of bombast, mined quotes, misrepresented physics, displays of gross (and wilful) ignorance, and so on, when they could have used all that time and effort far more productively, by developing BAC-like tests?

* though he came tantalisingly close, once
 
You must log in or register to reply here.

Back
Top Bottom