Arp objects, QSOs, Statistics

[Dancing David]

My obfuscation?

Just posting what I find. And care to comment on it?, or you going to ignore it, again?

And i'm quite sure BAC would be happy to point out any error he feels are in my posts, just as I would with him. We're not the same mind you know, we dont have to agree on everything!

Cease fire, this is an open thread and I would like to keep it that way.

 

[DeiRenDopa]

'Correcting' for the putative ejector, I get the following 'BAC probabilities':

a) Amaik: 0.011; Karlsson: 0.0034; 'regular': 0.0094

b) Amaik: 0.16; Karlsson: 9.2x10^-5; 'regular': 0.0017

Well. my recalculation using the correct Karlsson peaks matches your group a) 'BAC probabilities', but I am having problems with group b).

Hmm ...

Let's go over the calculation, with group b).

Here are the redshifts (from post #439):

b) 0.607041, 0.8798, 1.04486, 1.73016, 1.841, 1.9192, 2.57.

Next, correct for the redshift of NGC 5985 (0.0041); note that I followed your method for doing this Wrangler:

0.602931, 0.87569, 1.04075, 1.72605, 1.83689, 1.91509, 2.56589.

The 'nearest Karlsson peaks' are, respectively:

0.6, 0.96, 0.96, 1.96, 1.96, 1.96, 2.64.

That gives dz values of, respectively:

0.002931, 0.08431, 0.08075, 0.23395, 0.12311, 0.04491, 0.07411.

Multiplying these together gives 1.9x10^-9.

Taking r as 7, and n_k also as 7 (or did you use a different value, Wrangler?), the term ((2n_k)/3)^r is 48200.

Multiply 1.9x10^-9 by 48200 and you get 9.2x10^-5 (keeping only two significant figures).

Where does your calculation differ?

What is the point, however, if BAC won't comment?

It matters for many reasons. One of those reasons is that results should be independently verifiable (or reproducible) - if someone cannot get the same results, following the steps described, then whatever we're doing it cannot be called science, can it.

Why don't we just state our unanimous conclusion: the 'BAC probabilities' are just not giving us good, convincing statistical information.

Perhaps.

First though I'd like BAC to weigh in on the calculations.

After then (or anyway, if BAC continues to be silent) I'd like to look at what we've found, and see how (whether?) BAC's approach fails (and whether it could, even in principle, succeed ... under any circumstances).

 

[BeAChooser]

Originally Posted by BeAChooser
You're wrong. I'm willing to bet you that Arp et. al. cited every single case they could find where there were large numbers of quasers near galaxies and especially where those quasars seemed to align with some features of the galaxies. That's a random sample of all such such cases out there. And I took every one of those cases (that I could find) and calculated what the probability of finding redshifts that close to the Karlsson values would be, assuming that one looked at the entire population of such objects in the sky. And in almost all those cases the probability of seeing those cases was << 1. So you are mistaken. That is a random sample and the probability results should give you pause.

That is not a random sample.

It most certainly is a random sample among galaxies that were observed to have lots of quasars near them or have numerous quasars that seemed to be aligned with features of galaxies. As long as Arp didn't select only some from that group to publish (and I don't think he did), that's a random sample of that group. In which case, you are wrong, RC.

 

[BeAChooser]

Why don't we just state our unanimous conclusion: the 'BAC probabilities' are just not giving us good, convincing statistical information.

Wangler, please define Amaik and "regular" probabilities. In other words, what EXACTLY are we comparing to the probabilities I calculated?

 

[Reality Check]

It most certainly is a random sample among galaxies that were observed to have lots of quasars near them or have numerous quasars that seemed to be aligned with features of galaxies. As long as Arp didn't select only some from that group to publish (and I don't think he did), that's a random sample of that group. In which case, you are wrong, RC.

That is a random sample from a biased sample.

Back to your anology: They selected a biased sample which was all die showing the numbers in the set of numbers that that they were looking for. Then according to you they selected a random sample of this biased sample.

 

[Dancing David]

It most certainly is a random sample among galaxies that were observed to have lots of quasars near them or have numerous quasars that seemed to be aligned with features of galaxies. As long as Arp didn't select only some from that group to publish (and I don't think he did), that's a random sample of that group. In which case, you are wrong, RC.

Move goal posts much?

I laugh this time BAC! You don't have a clue, this time you are using words of a jargon in statistics and giving them your own special spin.

I know you can't admit it but you are flat out WRONG. That is the sad truth, and I am sure you can spin however you want.

You haven't addressed the issue of sample bias and error at all, which is even funnier, because that is the point of the thread.

So whatever, you can't answer a cogent argument with a cogent argument. You are clueless in the realm of statistics. No peer reviwed articles in population sampling, no textbook examples in population sampling. I can find hundreds and thousands to back my logic and you have your , well, spin I guess.

I really thought you had a chance on this one.

Whatever.

So pretend you are ignoring me, pretend that you have answered the arguemnt, you missed the boat.

 

[BeAChooser]

That is a random sample from a biased sample.

Not biased. Just the subset of the overall population being studied. The claim isn't being made that ALL quasars lie close to Karlsson values. Just that in those cases where there are inordinate numbers of quasars near galaxies, and particularly in cases where those quasars appear to be aligned relative to certain galactic features (such as the minor or major axis), the redshifts of the quasars seemed more prone to being close to Karlsson values than random chance would predict. You STILL don't seem to understand what I did in my calculation, RC.

 

[Reality Check]

Not biased. Just the subset of the overall population being studied. The claim isn't being made that ALL quasars lie close to Karlsson values. Just that in those cases where there are inordinate numbers of quasars near galaxies, and particularly in cases where those quasars appear to be aligned relative to certain galactic features (such as the minor or major axis), the redshifts of the quasars seemed more prone to being close to Karlsson values than random chance would predict. You STILL don't seem to understand what I did in my calculation, RC.

The objective of the statistics shoud be to distinguish between an association of QSOs with Arp objects and the association of QSOs with any other ojects. Just looking at Arp objects that have associations is a biased sample.

Back to your analogy: We want to see if the dice are biased to a certain set of numbers. As you have stated we need to calculate the statistics of a random sample of dice and see if the statistics match those of a set of fair dice or a set of biased dice.

Arp et. al. have not done this. They have looked at a biased sample (as you say "those cases where there are inordinate numbers of quasars near galaxies, and particularly in cases where those quasars appear to be aligned relative to certain galactic features (such as the minor or major axis)".

Where are the non-Arp objects in their papers?
If someone were to look at an equivalent set of non-Arp objects, scan through it to find a biased sample and and do your calculations then I woould expect them to get similiar probabilities. Does this mean that QSOs would then be associated withn non-Arp objects?

The fact of the matter is that Arp et al have not calculated any statistics at all. They have a number of examples and probability calculations. That is all.

 

[BeAChooser]

Just looking at Arp objects that have associations is a biased sample.

No it isn't. It's just the population of cases where there are associations between large numbers of quasars and galaxies or features of galaxies. Arp's theory is that in such cases the quasars came from the galaxy they are near. Arp's theory is that the redshift of the quasars decreases with time since creation. The question I'm asking is whether the redshift in those cases is unusually close to Karlsson values. That's not a biased sample. That's simply a sample (I maintain a random sample) from the population Arp was studying and theorizing about.

Look at it this way. RC, since you like analogies. Would I draw random sample from the entire human population if I was only interested in studying the temperature that people with a certain disease experience while sick? No, I'd take a random sample from the subset of people who experience the disease.

 

[Wangler]

Look at it this way. RC, since you like analogies. Would I draw random sample from the entire human population if I was only interested in studying the temperature that people with a certain disease experience while sick? No, I'd take a random sample from the subset of people who experience the disease.

But, if you wanted to ensure that the temperature you measure is related to the disease, you must also take a control sample from the general poplulation that is not infected with the disease.

If you didn't do that, then your sampling is incomplete. Which is what RC is trying to say, I think.

 

[Reality Check]

No it isn't. It's just the population of cases where there are associations between large numbers of quasars and galaxies or features of galaxies. Arp's theory is that in such cases the quasars came from the galaxy they are near. Arp's theory is that the redshift of the quasars decreases with time since creation. The question I'm asking is whether the redshift in those cases is unusually close to Karlsson values. That's not a biased sample. That's simply a sample (I maintain a random sample) from the population Arp was studying and theorizing about.

Look at it this way. RC, since you like analogies. Would I draw random sample from the entire human population if I was only interested in studying the temperature that people with a certain disease experience while sick? No, I'd take a random sample from the subset of people who experience the disease.

But you have to identify the "people with a certain disease". With people you can diagnose the disease but what if the disease was something that could be diagnosed by mistake?

However I suggest that you ignore these questions. I am more interested in your results for the other data sets in the thread. Any idea of when you will produce them or are you satisfied with the results form other people?

 

[DeiRenDopa]

Why don't we just state our unanimous conclusion: the 'BAC probabilities' are just not giving us good, convincing statistical information.

Wangler, please define Amaik and "regular" probabilities. In other words, what EXACTLY are we comparing to the probabilities I calculated?

Huh?

Perhaps BAC has me, DRD, on ignore?

Perhaps that's why he has not replied to any of my posts, requesting his direct inputs? Because he hasn't read them!

We could test this - well, others could - by one (or more) of you copying/quoting my posts (the ones in which I asked BAC direct questions).

BAC, in the chance that you can (and do) read this:

The "Amaik probabilities" are calculated using the same input data on redshifts as those used for calculating "Karlsson peak probabilities". The difference is that I created a set of "Amaik peaks", from certain inputs. These differ from the Karlsson peaks.

I created these as a test (albeit rather a weak one) of the significance (in the ordinary sense of the word, not statistical) of the Karlsson peaks - if they are some important feature of the redshifts of quasars (whether associated with galaxies or not, whether empirical or not), then one would expect that the "Amaik probabilities" would be (much) higher than the "Karlsson peak probabilities", and that there should be no relationship between them.

The "regulars" are another test of this same thing - they are derived by simply dividing up the range [0, 3] to produce 7 regularly-spaced "peaks" (oh, and I expressed all of them as having only two significant digits, so for example 1.13 rather than 1.125).

 

[Dancing David]

Not biased.

Oh, I see so in your gnomespeak, the word bias is statistics is not the convention of statistics, but only the meaning you impart to it. Wrong, it is a 'biased' sample.

Just the subset of the overall population being studied.

And what evidence is there that either of these two sets was chosen randomly, the burden is on you! They are not random by definition by the selection process engaged in. care to show that Arp used a random sample or just use your gnome of assertion and maintain that it is random without any evidence?

The claim isn't being made that ALL quasars lie close to Karlsson values. Just that in those cases where there are inordinate numbers of quasars near galaxies,

Which is undemonstrated since there is no control group of 'normative' distribution, the point of the thread which you ignore.

and particularly in cases where those quasars appear to be aligned relative to certain galactic features (such as the minor or major axis),

Another sample bias.

the redshifts of the quasars seemed more prone to being close to Karlsson values than random chance would predict. You STILL don't seem to understand what I did in my calculation, RC.

Yeah, you chose a biased sample from a biased sample preselected by Arp and then did a posteriory statistics on it.

So you engaged in pixie counting.

The resolution is is simple, take the normative sample.

 

[Dancing David]

No it isn't. It's just the population of cases where there are associations between large numbers of quasars and galaxies or features of galaxies.

And that means that it is not a random sample but a selected sample, which means that it is subject to 'bias' and 'error'.

Arp's theory is that in such cases the quasars came from the galaxy they are near. Arp's theory is that the redshift of the quasars decreases with time since creation. The question I'm asking is whether the redshift in those cases is unusually close to Karlsson values.

Which you can't even judge as a possibility without the normative distribution of the control group.

That's not a biased sample. That's simply a sample (I maintain a random sample) from the population Arp was studying and theorizing about.

And it is by defintion a 'biased sample', because it was preselected. It is not a random sample. The real problem is that there is no comparison to a control group.

Look at it this way. RC, since you like analogies. Would I draw random sample from the entire human population if I was only interested in studying the temperature that people with a certain disease experience while sick? No, I'd take a random sample from the subset of people who experience the disease.

No you wouldn't!

You already have the established baseline of 'normative' temperature to compare to your experimental sample.

And in fact most studies would aknowledge this and if they were using a data point that did not have an established normative distribution they would have to establish one, which as I have stated is the purpose of the thread.

In your analogy, the baseline temperature is not an established data point of comparison, because Arp did not establish that baseline!

Even by your own analogy, you are WRONG.

 

[Dancing David]

But, if you wanted to ensure that the temperature you measure is related to the disease, you must also take a control sample from the general poplulation that is not infected with the disease.

If you didn't do that, then your sampling is incomplete. Which is what RC is trying to say, I think.

Thank you Wangler, that is the point of the whole thread in the first place.

To point out that Arp did not use any controls, even though he still could.

 

[Dancing David]

DRD, BAC does not have you on ignore, they just feel that you are not speaking to the material that BAC wants to talk about.

That is part of the troll behavior of BAC, they do not respond to direct questions. They are ignoring you because it is not convnient to them to respond to you. By responding to you they know that they are likely to be caught out in logical errors and the inconsistancies of their alleged theories.

Just5 as BAC in the past stated that the flat rotation curve was imparted to galaxies during thier formation. that is why BAC is vague and opaque to criticism, they know that they are using bad reasoning and don't want to have it exposed.

 

[BeAChooser]

But, if you wanted to ensure that the temperature you measure is related to the disease, you must also take a control sample from the general poplulation that is not infected with the disease.

But what if it was a disease that affects everyone in that particular subgroup? In other words, perhaps there are no large collections of quasars associated with specific features that are unaffected. What then? Also, we are (or at least were at the time Arp et al started bringing these cases to our attention) dealing with relatively few cases. How do you get a control sample in that case?

If you didn't do that, then your sampling is incomplete. Which is what RC is trying to say, I think.

Well if you'd like, the control sample is what the mainstream claims the likelihoods should be given their observations of the number of quasars and their assumptions about the distribution of redshift.

Now, regarding my question to you earlier ... can you define "Amaik" and "regular"? What is the situation for the probability you are calculating in each case? If you don't know that, can you interpret their *meaning*?

 

[BeAChooser]

But you have to identify the "people with a certain disease".

The "disease" in this case is an assocation of a large number of quasars with a galaxy and/or features of the galaxy. That's how they identified it. Do you have any evidence that Arp, et. al., didn't report every case they knew about where there were large numbers of quasars or apparent associations of large number of quasars with minor axis or other features of galaxies?

 

[Dancing David]

But what if it was a disease that affects everyone in that particular subgroup? In other words, perhaps there are no large collections of quasars associated with specific features that are unaffected. What then? Also, we are (or at least were at the time Arp et al started bringing these cases to our attention) dealing with relatively few cases. How do you get a control sample in that case?

REally, you want to act that obtuse. I am shocked simply shockerd BAC!

The control sample is always going to be the same, normative distribution sample in the unaffedted population.

That is what this thread is about and you are still trying to tap dance around it.

the controls samples that i have proposed and that should have 1000 members or 10,000 even better would be:
-random spots on the sky and if you want random spots on teh sky not close to any galaxies , so two there
-'normal' galaxies
-AGN
-old galaxies
-young glaxies
-tidaly disrupted (this one will be a small sample) galaxies.

The same as I suggested early in the thread.

The point is, you can't claim a significan association unless you have a bseline level of the 'noise' of the system. Which is the reason I started this thread. Duh



Well if you'd like, the control sample is what the mainstream claims the likelihoods should be given their observations of the number of quasars and their assumptions about the distribution of redshift.

Now, regarding my question to you earlier ... can you define "Amaik" and "regular"? What is the situation for the probability you are calculating in each case? If you don't know that, can you interpret their *meaning*?[/QUOTE]

 

[Dancing David]

The "disease" in this case is an assocation of a large number of quasars with a galaxy and/or features of the galaxy. That's how they identified it. Do you have any evidence that Arp, et. al., didn't report every case they knew about where there were large numbers of quasars or apparent associations of large number of quasars with minor axis or other features of galaxies?

That is a red herring BAC!

The isue is that there is no control group with which to compare the the experimental group.

So your silly question is not that, the sample bias issue is not based upon if there is a complete survey of possible canidates but that there is nocontrol group to say that there is any meaning to the alleged association.

The evidence is that there is not a comparison control group.

taking an average density of QSOs/degrees of sky is not a control group.

So keep dancing to make Fred Astaire and Tommy Tunes envy you, you are dancing around the hole.

The problem is that you can't just assert that the average is the control group WHICH IS THE HUGE ERROR YOU MADE!

That is the assumption that is NOT made by mainstream cosmology, that QSOs are evenly spaced.

You keep digging that hole dude.

I have offered you teh solution and you igniore it, Arp , Burbidge et al. can make the control samples, the burden is on them and no one else.