Eric Lerner is presenting his recent paper on a static universe on PhysicsForums

[SelfSim]

Hmm .. perhaps somewhat naively, I made a request to him for a link to the source Galex dataset on the IS forum. (He may have already referenced the source in his paper, but I've kinda lost the trail of it all during the break in conversation).

 

[ianw16]

Hmm .. perhaps somewhat naively, I made a request to him for a link to the source Galex dataset on the IS forum. (He may have already referenced the source in his paper, but I've kinda lost the trail of it all during the break in conversation).

The Lerner 2018 paper references the Lerner et al 2014 paper regarding the dataset. The 2014 paper has no reference in the references, and the only mention in the body of the paper is to GALEX MIS3-SDSSDR5. This is the Medium Imaging Survey. There is a little on this here:
Chapter 2 - GALEX Science Surveys

There is also a paper giving details on the GR2 & GR3 data releases, which I presume MIS3 is a part of:

THE CALIBRATION AND DATA PRODUCTS OF GALEX
Morrisey, P. et al. (2007)
http://iopscience.iop.org/article/10.1086/520512/pdf

The above paper is one of a number from GALEX in a free issue of ApJ, in December 2007. The articles can be accessed here:

The Astrophysical Journal Supplement Series, Volume 173, Number 2, 2007 December - IOPscience

 

[sjastro]

I was able to dig up a Galex image data based on Lerner's response at IS.
While it does not pertain directly to Lerner's data it is an example of the Medium Imaging Survey and serves to highlight the image quality.

This a single image of the M31 region and the grainy background indicates the image is extremely noisy as confirmed by the Pixinsight measurement of σ = 1.358 x 10 ̄ ².
The FWHM using the same software was 1.75 +/- 0.31 arcsec.
Using AIP4WIN software and handpicking six stars the FWHM was 7.19 +/- 0.56 arsec.

Pixinsight gave a low value as it mistook the background noise as stars while the AIP4WIN results are higher than being in the vicinity of 5.0 arcsec as specified by Galex.
In both cases the FWHMs are rejected due to the high background noise due to the difficulty of accurately determining a radius value in order to calculate the FWHMs.
This would also apply to Lerner's radius measurements.

In order to improve the S/N ratio the Galex images would need to be stacked much like the Hubble image data (although the drizzling/dithering process will not work on Galex images).

The question that needs to be asked is whether Lerner performed his radius measurements on single image data or stacked data?

It should be noted however that irrespective of how Lerner performed the measurements his cutoff results are around 50% of the specified FWHM instead of being the lower limit.

 

[SelfSim]

Posted the above question but it seems he didn't answer it (ie: the 'stacked' or 'not stacked').

... It should be noted however that irrespective of how Lerner performed the measurements his cutoff results are around 50% of the specified FWHM instead of being the lower limit.

I have said before that the difference between FWHM and radius is just the same as diameter and radius--a factor of 2. That is where your 50% comes from. Every galaxy was measured individually.

.. and to be fair he has made this same response 3 times, now .. but I'm not at all sure this answers the last question which was about which FWHM value he chose to use (and why), given that the entire point of the lead up to the question was that one can select different (max, min, etc) values of FWHM for his cutoff .. So, I interpret this as being independent of whether he uses radius or diameter (as long as he's being consistent about it)?

 

[sjastro]

Posted the above question but it seems he didn't answer it (ie: the 'stacked' or 'not stacked').
.. and to be fair he has made this same response 3 times, now .. but I'm not at all sure this answers the last question which was about which FWHM value he chose to use (and why), given that the entire point of the lead up to the question was that one can select different (max, min, etc) values of FWHM for his cutoff .. So, I interpret this as being independent of whether he uses radius or diameter (as long as he's being consistent about it)?

Repeating the same point three times doesn't make it right.
The FWHM does represent a resolution limit.
One can argue he should multiply his radius values by two so one is comparing apples with apples.
Irrespective if one uses radius or diameter there is a problem that shows up particularly in the NUV data.
The average radius is 2.6 arcsec which is less than 50% off the specified FWHM of 5.3 arcsec.
Lerner is somehow able to measure a radius that is below the resolution limit of Galex.

Then there is the issue of the Hubble data where this 50% value applies to all filters.
As a matter of interest here is the FWHM performance for Hubble's WPFC-3 camera with various filters.
Hubble have not only factored in the wavelength but also transmittance which also affects FWHM.

Ignoring the individual versus stacked images issue might lead to a tap on the shoulder from the moderators....

 

[SelfSim]

Ok .. re-reading post#81 here and his subsequent responses, I get the impression that either:

i) he genuinely doesn't understand the point we're making or;
ii) he's deliberately obfuscating the matter with his 'two clear peaks for point sources and extended ones' argument and he's using the 'radius/diameter accounts for the difference', as a smokescreen cover-up?

Are we in synch with that take on the situation?

If so, I suggest we revisit the issue later and for the moment, we see if we can get a straight answer on the 'stacking' question(?)

 

[sjastro]

Ok .. re-reading post#81 here and his subsequent responses, I get the impression that either:

i) he genuinely doesn't understand the point we're making or;
ii) he's deliberately obfuscating the matter with his 'two clear peaks for point sources and extended ones' argument and he's using the 'radius/diameter accounts for the difference', as a smokescreen cover-up?

Are we in synch with that take on the situation?

If so, I suggest we revisit the issue later and for the moment, we see if we can get a straight answer on the 'stacking' question(?)

I have found with every image file from Galex comes a corresponding sky background file.
These files are stacked or more precisely the sky background is subtracted from the image file.
This reduces the noise value somewhat and gives more "consistent" results when using Pixinsight or AIP4WIN.
Pixinsight FWHM:- 7.05 +/- 1.93 arcsec.
AIP4WIN FWHM:- 7.35 +/- 0.02 arcsec.

The FWHM is still out as the noise value remains high enough to distort the values.
A more precise question to Lerner is if he sky background subtracted the data which could impact on the cut off values.

 

[SelfSim]

I have found with every image file from Galex comes a corresponding sky background file.
These files are stacked or more precisely the sky background is subtracted from the image file.

And yet Lerner insists that the images he used are not 'stacked':

My answer had no ambiguity at all. GALEX images and HUDF images of individual galaxies were measured. That is the data used. The images are not stacked. Stacking is the process where many images from different galaxies are added together. We measured each galaxy separately. The galaxy measurements were then combined into samples as described in detail in the papers and a median radius for each sample determined.

So whilst his image files may not be stacked, the sky background files should have been used to subtract the background noise from these (ie: by using 'stacked' sky background files)?

How he can then claim that he's measured accurate radii (within the FWHM spec of the scope) escapes me(?) ..

A more precise question to Lerner is if he sky background subtracted the data which could impact on the cut off values.

Yep .. I agree (as per the above).
Sheesh! .. He's not very forthcoming in volunteering info thatshould be easy for him to volunteer!?

 

[SelfSim]

Hmm .. question put to him .. (he's also getting grumpy).

 

[sjastro]

And yet Lerner insists that the images he used are not 'stacked':So whilst his image files may not be stacked, the sky background files should have been used to subtract the background noise from these (ie: by using 'stacked' sky background files)?

How he can then claim that he's measured accurate radii (within the FWHM spec of the scope) escapes me(?) ..

Yep .. I agree (as per the above).
Sheesh! .. He's not very forthcoming in volunteering info that should be easy for him to volunteer!?

A fundamental principle of astronomical photometry is obtaining a high S/N ratio.

In the case of Galex extraneous noise can be an issue particularly for fainter galaxies and images need to have the sky background subtracted.

Users of GALEX data should be aware that the sky background in GALEX images is in general not a constant across an individual field, particularly in the FUV. The main contributor to the FUV sky background is UV light from hot stars in the Milky Way disk scattering off dust cirrus. There is structure in the cirrus on a wide a variety of scales. Even at high Galactic latitudes, there are areas of the sky where there is significant cirrus structure in the FUV background. Users interested in faint diffuse light in the FUV, such as in low surface brightness galaxies or in the extended disks of spiral galaxies, should be aware of the potential confusion with the Galactic cirrus.

If he hasn't subtracted the sky background the image data is useless.

In case of Hubble, Lerner is in for a nasty shock as the HUDF image filter data is classified as level 2 or 3.

There are several levels of data in the HLA:

• Level 1 data are the individual exposures, projected ("drizzled") onto a common frame for each visit.
• Level 2 data are the combined images (exposures with the same filter, same camera, and within the same visit).
• Level 3 images are deep combined images or mosaics creating by combining data taken in different visits to the target.
• Level 4 data are false color images (a combination of two or three colors, depending on what is available). They are included with other selections to help users assess data quality.
• Level 5 data are the community-contributed high-level science products (HLSP).
• Level 0 (calibrated, unprocessed) are not available for direct searches but may often by found using the More... link.

Assuming Lerner has accessed the HUDF data from the Hubble legacy site he has been using combined (stacked) image data without knowing it.

It seems to me Lerner is now completely on the defensive.

 

[SelfSim]

It seems to me Lerner is now completely on the defensive.

I agree ... I think his last post was more or less a signoff one too.
I don't get a strong impression that Lerner did the analysis himself because he doesn't seem to be able to explain it, he doesn't appear to understand the data they've used and he is evasive in his responses to questions about it.

There's something quite odd about this whole paper (not the least of which is how it actually got published in MRAS).

I think its time to summarise the key failings in this paper (and share them around).

 

[sjastro]

I agree ... I think his last post was more or less a signoff one too.
I don't get a strong impression that Lerner did the analysis himself because he doesn't seem to be able to explain it, he doesn't appear to understand the data they've used and he is evasive in his responses to questions about it.

There's something quite odd about this whole paper (not the least of which is how it actually got published in MRAS).

I think its time to summarise the key failings in this paper (and share them around).

I'd give Lerner a few days grace before declaring he has signed off.
I've noticed the FB Astrostatistics page on the subject is now closed.
This comment grabbed my attention.

Lerner was invited to visit ESO about 10 years ago or so where he first wrote some papers on the Tolman Surface Brightness test in a rather embarrassing fashion that succumbed to many of the same problems that Johannes Buchner outlined (forgive me if I do not link to it). Every few years, he pops his head up hoping someone will take notice. Few do because, well, he hasn't kept apace of actual history of how cosmology has developed on these many decades. And so here we are.

I blame the editor of MNRAS and the referee of the paper for the embarrassing acceptance. I suggest writing them a letter letting them know that they messed up.

As an interesting side story a few years ago ESO "peer reviewed" a mathematical function I came up with that was able to extract far fainter detail from ESO images than their own processing.
Like the questions posed to Lerner in this thread I had to answer ESO's questions regarding stacking images, sky background subtraction, signal/noise ratio of the background etc.
It brought back some memories.........

 

[FrumiousBandersnatch]

... The images are not stacked. Stacking is the process where many images from different galaxies are added together. We measured each galaxy separately.

That doesn't sound right (my bolding); how would that work?

 

[SelfSim]

That doesn't sound right (my bolding); how would that work?

Hi FB;

His response seems typical of his style to me. Ie: ambiguous, unclear and confusing.

I assumed he meant multiple images of a given galaxy, (same filter and camera), then repeat the same process for other target galaxies ... (otherwise it wouldn't make any sense).

However, it could also be interpreted as just being a totally wrong method.

Either way, our point is that the publically available HUDF images are combined images (as sjastro points out in post#70), and are thus operating close to theoretical Rayleigh resolution limits .. (which means the filters used impact the validity of his method), so if he's directly comparing those with Galex images of the same targets, (where the Galex images haven't had the sky background subtracted), then its not an apples-apples comparison and his 'method' needs to address this (it doesn't appear to).

How he's processed the images is key and he hasn't been very forthcoming with info about any of that, nor has he helped any on tracking down the specific data he used.

 

[FrumiousBandersnatch]

Hi FB;

His response seems typical of his style to me. Ie: ambiguous, unclear and confusing.

I assumed he meant multiple images of a given galaxy, (same filter and camera), then repeat the same process for other target galaxies ... (otherwise it wouldn't make any sense).

However, it could also be interpreted as just being a totally wrong method.

Either way, our point is that the publically available HUDF images are combined images (as sjastro points out in post#70), and are thus operating close to theoretical Rayleigh resolution limits .. (which means the filters used impact the validity of his method), so if he's directly comparing those with Galex images of the same targets, (where the Galex images haven't had the sky background subtracted), then its not an apples-apples comparison and his 'method' needs to address this (it doesn't appear to).

How he's processed the images is key and he hasn't been very forthcoming with info about any of that, nor has he helped any on tracking down the specific data he used.

OK, thanks - probably just finger trouble; but it seemed a bit odd.

 

[SelfSim]

OK, thanks - probably just finger trouble; but it seemed a bit odd.

Not altogether sure its just (his) finger trouble either ...

 

[sjastro]

Lerner's latest responses are not responses at all.

Jean Tate--the stellarity index has the name "NUV_class_star" and "FUV_class_star"
Selfsim--the HUDF data was from Coe D., Bentez N. Sanchez S. F., Jee M., Bouwens R., Ford H. 2006, AJ 132, 926
Jean Tate--they key thing is that bolometric units are just energy. AB units are energy per unit frequency, so since the light is redshifted, that means different exponents for the surface brightness. No physical difference, just how you measure it.

He has avoided answering the question if the Galex image data had the sky background subtracted while his reference to Coe et al. paper tells us nothing about specific details of the HUDF image data.

I think he has been well and truly caught out particularly on the Galex issue.
It seems just about everyone except for the reviewers at MNRAS have found problems in Lerner's paper.