And this is a bad thing? Why should anyone champion scientific certainty?
Should we just delete the millions being spent on gravity wave research because we are certain Einstein is right anyway?

This paper I just posted, seconds after Parejk.. posted the same reference, is new as of that day, and has subject matter exactly pertinent to this thread. If the information would have been different, but still relevant, I might have posted it as well.

I don't have a 'preferred' value for Ho - be it 7 or 7000 - it is a size parameter. What I worry about is artificial constraints drawn not from observations, but expectations. Schaefer has the same concern, and has articulated this concern in a statistical study of Ho methodology. Schaefer has concluded there is a bandwagon effect evident in the reduced data that is not scientific. Is he right?

__________________
jwj

It's a big universe out there...is it really unwinding, really burning out?

Well, the fact that the distribution of the best values does not appear to be non-Gaussian is counter-evidence against his claim that there is a bandwagon effect in the papers. Why would it be evident only in the errors and not in the best values themselves? Doesn't make sense to me.

I am not sure what "underweighed" is supposed to mean here.

Are you aware that such a selection effect would lead the derived H0 to be biased low, not high? Besides, you have not offered any evidence for it.

You would get the most accurate answer for the distance to cluster centers, but not for H0. For H0 estimation, the difference is less than 1 km/s/Mpc. See above. You have provided no evidence that would refute this.

Data is always limited in the real world. But even if 100 SBF measurements could be done, it would probably still be most effective to do only one measurement per cluster because 100 different cluster and flow environments would thus be sampled and any unaccounted for systematics specific to these environments would have minimal weight. I want to stress the importance of this point. 100 measurements taken from a single cluster would all be subject to the same local systematics and H0 thus derived could be wildly off the mark.

It is not "my method", just the standard method in the business, used by HKP and others.

You are missing the point. The point was not to assess the combined effect of measurement errors and galaxy locations on H0 but to see the effect from galaxy locations only.

This is too vague to say anything about. You first need to provide these details: galaxy ID, distance, errors and the reference used.

No, because a) clusters do not have a diameter of even 12 Mpc on average and b) the number density of galaxies in clusters from the center to the edge falls off more rapidly than a Gaussian curve. So far you have not even acknowledged the fact that the number density actually falls off significantly toward the edge.

They are not distances to "some galaxies", they are all distances to Virgo itself. You can't use these figures to estimate the Virgo diameter, only the distance.

It is not a theoretical calculation. It comes straight from the measured diameter and cluster distance.

Wrong.

I bet this is from Mei 2007, ApJ, 655, 144. Care to elaborate on what she says about the lengths of these axes?

Hi Zahl,

I am travelling on the road for a number of days, so can't get to all of your points immediately. But the following stood out as a no-brainer.

I stated that:

Quote:
So your theoretical calculation gave less than 3 Mpc

You replied:

The measured diameter you referred to was 11 degrees, which you used to calculate the diameter of the Virgo Cluster, assuming it is spherical in shape.
Diam = distance*theta*pi/180 = 15.8*11*3.1416/180 = 3


If you look at Table 2, of the "Final Results ..........." paper by Freedman, et al that we have been discussing, there are 6 Cepheid galaxies listed from Virgo. Of these, NGC 4321 is located at RA 185.73, Dec 15.82 and Ngc 4536 is located at RA 188.61 and Dec 2.19 (values in deg). These are taken from a list of only 6 of the 2000+ galaxies in Virgo.

The central angle between these two galaxies is no less than the difference in declinations, 15.82 - 2.19 = 13.6 deg. Accounting for the difference in Right Ascension increases this angle. So how can you say that the measured diameter of the Virgo Cluster is 11 degrees?

When considering all the members of this cluster, I am willing to speculate that the size of the cluster is much larger.

The standard deviation in the best value after 2001 is much tighter than it was before; that won't show up unless you plot both data sets.
Whenever you talk about a bandwagon effect, you also have to take into account those who do not jump on.

For example, Efforts to remove systemics from Tully-Fisher analysis are not weighted the same by astrophysicist as papers that produce the consensus in the Hubble value. While there may be TF systemics that are biasing the data, it could be this worrysome bandwagon effect.

Don't care. If there is a potential for selection bias, the error bands should reflect this. Neried seems to think selection bias is highly unlikely, and this is one of her strong areas.

__________________
jwj

It's a big universe out there...is it really unwinding, really burning out?

Quote: TomT
So your theoretical calculation gave less than 3 Mpc

More on this. Here is a link to a study of galaxy distances.

http://arxiv.org/ftp/astro-ph/papers/0503/0503440.pdf

Table III gives distances to 17 spiral galaxies in the Virgo Cluster determined using the Tully Fisher method (type dependent) and calibrated using the HKP Cepheids. Note the distance of 28.1 Mpc to NGC 4343 and 12.6 to NGC 4569. The difference of 15.5 Mpc is approximately the distance between 2 galaxies within Virgo (the actual distance is slightly larger when accounting for the Right Ascension and Declination of each galaxy). This distance is far greater than "a couple of Mpc". So now we have reviewed a total of only about 20 of the 2000+ galaxies in Virgo, and found the cluster size to be at least 15.5 Mpc. This should suffice to illustrate the point.

The angle is cos^-1(cos(2.88°)*cos(15.82°)*cos(2.19°)+sin(15.82°)*sin (2.19°))=13.9°. I already gave you the source for that 10 degrees, but here it is again: NASA/IPAC Extragalactic Database.

http://nedwww.ipac.caltech.edu/cgi-b...&img_stamp=YES

NGC 4536 belongs to the Virgo II galaxy cloud, also called Southern Extension, a structure located to the south of the main Virgo cluster and it is debatable if it is a true member of the Virgo cluster. See 2001A&A...375..770F.

The calculation I gave earlier assumed an average cluster diameter of 12 Mpc or 43 degrees in case of the Virgo cluster. I don't think you can refute that. To validate dgruss' ATM idea of 30 Mpc cluster depths you would need to demonstrate an angle of more than 100 degrees between two members of the Virgo cluster.

?? That the standard deviation is tighter has no relevance to what I wrote above. The question was why is the distribution of the best values Gaussian if there is a bandwagon effect in the papers (as demonstrated by the alleged non-Gaussianity in the errors)?

I'm still not sure what you meant by 'underweighed' when you wrote that "the more recent methods that incorporate metallicity in the determination of the Cephied distances are being 'underweighed'". I think you are just spreading UNDO (uncertainty & doubt) towards the mainstream with little to back it up.

Are you saying that you have no evidence for selection bias?

Huh? There are no such galaxies anywhere in that paper. Your figures are useless anyway as you don't quote the errors. Please don't link to any papers that don't even give measurement errors, I'm tired of that unscientific nonsense.

OK, you said 10 degrees not 11. And, as you point out, the 2 Cepheids in Virgo are 13.9 degrees apart. So are you still arguing that the diameter of Virgo is 10 degrees?

What 2 Cepheids are you talking about?

The calculation I gave earlier assumed an average cluster diameter of 12 Mpc or 43 degrees in case of the Virgo cluster. I don't think you can refute that.

You're joking, right? Read my post #244 above.

Oooops. I gave the wrong link. Here is the correct one.

http://arxiv.org/ftp/astro-ph/papers/0408/0408348.pdf

Repeating my point:
Table III gives distances to 17 spiral galaxies in the Virgo Cluster determined using the Tully Fisher method (type dependent) and calibrated using the HKP Cepheids. The Table III values are from the Tully Fisher equations (6) and (7) which contain the error component.
From these are calculated distances of 28.1 Mpc to NGC 4343 and 12.6 to NGC 4569. The distances with error bars are 28.1 +/- 2.1 Mpc for NGC 4343, and 12.6 +/- 1.1 Mpc for NGC 4569. The difference of 15.5 (13.4 to 18.6) Mpc is approximately the distance between 2 galaxies within Virgo (the actual distance is slightly larger when accounting for the Right Ascension and Declination of each galaxy). This distance is far greater than "a couple of Mpc". So now we have reviewed a total of only about 20 of the 2000+ galaxies in Virgo, and found the cluster size to be at least 15.5 Mpc. This should suffice to illustrate the point.

I just wondered why you were suddenly claiming that I had calculated an angle between "2 Cepheids" when it really was two galaxies.

I haven't checked the errors you give against that paper, but they can not be correct. Even in the Tully & Pierce paper that dgruss has promoted in this thread it says that individual galaxies have a distance error of 15-20% in the Tully-Fisher method. Besides, The B-band TFR used in the linked paper has significantly more scatter than the I-band relation, resulting in larger distance errors.

OK, so now we are completely clear on the point that the calculation is for two cepheid galaxies in Virgo. So let's settle this simple, obvious issue.
Your source claims the measured diameter of the Virgo cluster is 10 degrees.
I pointed out that six galaxies in Virgo have Cepheids that were listed in the "Final Results ...." paper by Freedman et al. I provided Right Ascension and Declination data for two of these galaxies. You yourself calculated the exact angle between the two galaxies as 13.9 degrees.
(1) So can we first agree that 13.9 is larger than 10, so the diameter of the Virgo Cluster has to be greater than the 10 degrees that your source claims?

Now the distance between these same 2 galaxies can be calculated using their RA and DEC in conjunction with their distances. The result is 3.7 Mpc which obviously is greater than the 2.7 Mpc calculated from your 10 degree claim for the diameter of the cluster, and the "couple of Mpc" you claimed earlier. Note that the actual distance between the galaxies is less than that calculated from the central angle of 13.9 deg and the distance to the cluster center (15.8 Mpc) because these 2 galaxies are located in the area of the cluster nearer to us than the center (14.9 and 15.2 Mpc).
(2) So can we at least agree that the diameter of the Virgo cluster has to be at least 3.7 Mpc.
Further, this is based on only this sample of 6 galaxies in the "Final Results...." paper by Freedmann et al, of which we used these two. And this is only 6 galaxies of the 2000+ in the Virgo Cluster?

No, you have only demonstrate that two galaxies are more than 30 Mpc apart. The angle is dependent on the location of the observation, and the cluster does not have to be spherical. It could be elongated more in one direction than another, as it is believed Virgo is.

It depends on the membership of NGC 4536 which is largely a matter of subjective definition. 2001A&A...375..770F did some detailed analysis and couldn't give a robust answer. NED is a reliable source and according to it that's not the case. But this is all really beside the point.

What is the error on that 3.7 Mpc figure? And I have never claimed that the Virgo cluster diameter is a "couple of Mpc". Earlier in this thread I wrote that the characteristic radius of clusters can be up to 5 Mpc and I cautiously used an average diameter of 12 Mpc in my H0 error calculation. This is really the point of this discussion if we want to assess the reliability of HKP's SBF determination of H0 based on 6 galaxies.

Not yet. Consider the following from Professor Barbara Ryden (Department of Astronomy, The Ohio State University): "Poor clusters of galaxies (of which the Local Group is one) contain less than a thousand galaxies. Most clusters are poor. However, there are also a few rich clusters, containing over a thousand galaxies. Where poor clusters are only 1 megaparsec (Mpc) across or less, a rich cluster will be about 3 Mpc to 10 Mpc in size.

The Virgo cluster (seen in the image below) is an example of a moderately rich cluster of galaxies. It contains over 2000 galaxies, including the giant elliptical galaxy M87. The Virgo cluster is approximately 15 Mpc away and 3 Mpc in diameter. Since the Virgo cluster is the nearest rich cluster to us, it occupies a large region of the sky, about 10 degrees across."

http://www.astronomy.ohio-state.edu/...8/notes34.html

So there you have it. The Virgo cluster has a diameter of 3 Mpc and is richer and more extended than most clusters. Cluster sizes go up to 10 Mpc, according to Professor Ryden.

Then all the good professor has left to do is provide us with the names of those 2000 galaxies, and we can check the right ascension and declination of each and make a preliminary conclusion if he is right. We will then need some selected distances to wrap up the investigation.

What I wrote was in the context of angles. It is obvious that you can't demonstrate that he was right by using only angles as you did in post #244 to which that quote replied.

You can start by scouring through the New General Catalogue and report what you find. Just don't forget the errors. Professor Ryden is a woman, btw.

Here:

http://nedwww.ipac.caltech.edu/level...D/LMC_ref.html

Is an updated compilation of LMC distances put together by Steer & Madore.

It is easy to see the trend (See Thumbnail) identified by Schaefter (//xxx.lanl.gov/abs/0709.4531) towards both smaller errors and tighter estimates between 2000 and 2006. Notice that the three latest papers buck the trend. The two highest outliers (~19.2 and 18.8) are the Tully-Fisher distance and a Baynsian "quasi-geometric" Cephied analsys by Barnes et al.

http://articles.adsabs.harvard.edu/c...&filetype=.pdf

The Barnes' Cephied distance does not include metallicity, which would only make the outlier worse.

Schaefter looked at the trend, then looked for explanations. He ruled out advances in methodology, demostrated that the trend tightening is unnaturally statistically significant, and noted that the same trend is not seen in distance estimates to other objects. He drew the conclusion somebody is likely pencil sharpening, and the issue needs to be addressed.

(Note: Not all the methods included in S&M include straight-forward error estimates, and those that do not have been included with the error tick left off.)

There was a lot of confidence in 2000. We were entering the era of precision cosmology; and there are valid reasons for error estimates to shrink and values to converge. Since then, (and very unexpectedly) the error bars on the estimates of distant supernova magnitudes have opened up a lot. Don't expect this issue to go away: We don't have a good handle on the distance to our nearest neighbor.

Attached Images

LMC Distances.GIF (9.0 KB, 11 views)

__________________
jwj

It's a big universe out there...is it really unwinding, really burning out?

Let's take a second look at this paper. http://arxiv.org/abs/0709.4531

Schaefer's argument was that the results from 31 post-2002 papers with LMC distance determinations were too tightly concentrated around the HKP best value of 18.50 for the distance modulus and that this was likely to be due to a bandwagon effect. As evidence he gave a cumulative distribution plot of the quoted and theoretical Gaussian errors and found that the deviation between the two was unlikely at the 0.0023 probability level according to the Kolmogorov-Smirnov test.

Here's a q-q plot of the errors [(μ-18.50)/σ] from the 31 post-2002 papers (Schaefer's Table 1):



There's clearly some excess kurtosis in there, but the evidence is not as strong as Schaefer got from Kolmogorov-Smirnov, being significant only at the two sigma level or so. D'Agostino omnibus normality test gives P=0.054, Shapiro-Wilk gives P=0.025 and Jarque-Bera P=0.035.

Here's that same plot, but this time with the random errors only:



All evidence of non-Gaussianity has suddenly vanished. D'Agostino omnibus normality test P=0.70, Shapiro-Wilk P=0.34 and Jarque-Bera P=0.80. As the distributions of best values and random errors appear to be just fine, it seems that the problem lies with the systematic errors. More about this later.

Schaefer looked for improvements in methodology that could explain the systemics, but he found none. He also looked for similar trends in distances to objects not highlighted in the HKP, and did not find them. All of these analysis involve small numbers and perhaps too many degrees of freedom to draw hard conclusions. But I think it is naive to pretend all research groups are not influenced by 'industry wide' trends while they are constraining their new analysis.

I have been involved in a number of industry-wide blind studies; and if the study is truly blind, a shotgun pattern emerges. Almost without exception, when samples are reanalysed after the first round of results has been released, the 'reanalyzed' data closes around the mean or mode value derived in the first round.

Personally I learned that if you use a new pipet in the first round, you spend less time explaining your data the second time around. The results will also 'improve' if I spend more or less time waiting for the pipet to drain before touching it off, depending upon which side of the mean my analysis ended up on. But that would be a bandwagon effect.

__________________
jwj

It's a big universe out there...is it really unwinding, really burning out?