Facts against BB Theory

jacob navia writes:
Le 28/04/2014 22:13, wlandsman a ecrit :
For example, Sutter et al. (2013,http://arxiv.org/abs/1310.7155) find
nearly 1000 voids in the Sloan survey, and compare these with their
cosmological simulations, and conclude: "Our void abundances,
ellipticity distributions, and radial profiles all indicate that voids
in theory have the same sizes, shapes, and interior contents as
observed voids."

The second part (that you make reference as a proof for BB theory) is
that the catalog is very similar to orthodox cosmology simulations
realized in software.

I wouldn't say "proof"--agreement supports BB cosmology, it doesn't
prove it.

But let's review what they are doing. Sutter et al. want to compare
observations to models. The observations are taken with a particular
instrument, at a particular location, covering a particular region of
the sky. To compare the models with the observations, they want to
replicate the effects of observing their simulated data in the same
way the real observations are taken. This is the function of the
"survey mask", which accounts for the geometrical and other effects of
the actual observations.

And this is where I disagree with the author of the paper. I read
paragraphs like this, for instance:

quote
In Figure 7 we compare the number function of voids in the CMASS Mid
data sample to all our mocks. First, the unmasked N-body Mock simulation
hosts roughly three times as many voids per unit volume than the data,
even though they have similar galaxy populations. This occurs at all
scales, though there are approx 4 times as many small voids in the
unmasked mock as in the data.
end quote

So, the raw data doesn't at all fit with the simulation. Then, he
applies his mask, what is probably OK if the mask used is the SAME as
the mask he used with his data. But I did not find that sentence in the
paper and can't be sure of that fundamental fact.

They don't apply a mask to the data--the mask is *derived* from the
geometry and detector effects of the actual observations, so that
the simulation can be made comparable to the observations. There's
no a priori reason to expect the unmasked simulation to look like
the observations, as the masking is necessary to account for the
details of the observations.

As the text says, "we apply the same survey geometry to mock galaxies
as is used in the SDSS DR9 samples" [...] "We project all galaxies
onto the sky and apply the mask in Figure 1." (where figure 1 is the
actual DR9 survey area). So they do say that the mask they are
applying is directly derived from the actual geometry of the survey
sample, which is what you'd expect in order to make the simulations
comparable to the observations.

Besides that, a few lines below, the paper says...
quote
As found in Sutter et al. (2013), the best match to voids in low-density
galaxy surveys comes from adjusting the "void parameter" Dv to -0.015.
end quote

There we start massaging the data so that it fits whatever we want it to
fit, sorry. Why is the simulation a low density survey? Maybe I did not
understand everything but the article is vague here.

Previous (and first) sentence in that paragraph: "In the same figure
we plot the theoretical number function of Sheth & van de Weygaert
(2004), which was derived from an excursion set formalism."

At this point, they are a comparing a theoretical model (the black line
in fig. 7) with their unmasked simulation, so there's actually no data
involved in this comparison. The "void parameter" is a parameter of the
theoretical model in the Sheth & van de Weygaert paper.

Let's go on:

quote
While the number function roughly agrees with the order of magnitude of
the full N-body Mock void population, it overestimates the number of
voids in all size ranges considered here. It also does not fall off as
steeply as in the mocks, though this might be influenced by
finite-volume effects.
end quote

So, he must acknowledge that the data just DOESN'T FIT!

Still comparing their simulation to the S&vdW theoretical model, no
data involved. What they are saying is that the S&vdW theoretical
model isn't a great quantitative fit to their simulation.

Next sentence after acknowledging that?

quote
Still, the correspondence of these curves shows that theoretical
modeling can qualitatively match unmasked void populations, but further
adjustments must be made to match void statistics from masked volumes.
end quote

Excuse but there is NO WAY that THIS can be presented as an exact
correpondence between the data and the BB models!!!
The two sentences: one that says the data doesn't fit, and the very next
one where he concludes that the survey supports BB theory just DO NOT
MAKE SENSE!

Still no data involved. Still comparing theory to simulations.

All they are saying here is that the S&vdW theoretical model has some
qualitative agreements with their simulations, so it may have some
merit as a theoretical model, but more work is needed just to make the
theoretical model comparable to the actual observations.

The meat of the paper is in the comparison of the masked "mock data"
to the observations. Those do appear to be in excellent agreement.

-dan

In article , Jos Bergervoet
writes:

In 1910 Newtonian gravitation had been
around for a long time. Ask yourself a
simple question: Did Newtonian theory
predict that matter would be organized
into "island universes" we call galaxies?

No! Not a chance!

The fact is that no-one had calculated it. Such computing power did not
become available until the late 1970s.

While a prediction is more emotionally satisfying than a postdiction,
one shouldn't put too much weight on historical contingency which
determined whether theory or observation was first.

There are "dictions" with several levels of strength:

Indeed.

1) The theory predicts it and later it is observed.
2) The theory *predated* the observation but only
afterwards the postdiction is made (case above!)
3) The theory is formulated *after* the observation
and is able to postdict the observation without
effort.
4) The theory predated the observation but needs to
be extensively fitted to match one new observation.
5) The theory is formulated after the observation and
still needs to be extensively fitted to match it.

It sounds reasonable to say that cases 1 and 2 are
actually equivalent

I agree.

(and 4 and 5 are suspicious..)

I agree.

One could also argue that 3 is equivalent to 1 and 2, at least if it is
clear that observations are not used to shape the theory and/or that the
theorists didn't know about the observations.

Some confusion arises in debates because of failure to distinguish 1 and
2 (and perhaps 3) from 4 and 5.

On Friday, May 2, 2014 4:23:43 AM UTC-4, Phillip Helbig---undress to reply wrote:
In article , "Robert L.
Oldershaw" writes:
In 1910 Newtonian gravitation had been
around for a long time. Ask yourself a
simple question: Did Newtonian theory
predict that matter would be organized
into "island universes" we call galaxies?

No! Not a chance!

The fact is that no-one had calculated it. Such computing power did not
become available until the late 1970s.

While a prediction is more emotionally satisfying than a postdiction,
one shouldn't put too much weight on historical contingency which
determined whether theory or observation was first.

The implication that galactic structure
would have been definitively predicted if
only they had enough "computing power" in
1910 seems unlikely in the extreme to me.

[Mod. note: any argument behind that conclusion? -- mjh]

The evaluation of the role and importance
of predictions vs retrodictions given above
does not correspond to any explanation of
how science is supposed to work that I could
support.

I am concerned for the health and progress
of science, but perhaps we are well beyond
the point of concern. The fact that so few
seem to share these concerns is puzzling and
does not bode well for the immediate future
of physics. Still, I think science is self-
correcting in the long run so there always
reason for optimism.

[Mod. note: reformatted -- mjh]

On Saturday, May 3, 2014 2:32:27 AM UTC-4, Phillip Helbig---undress to reply wrote:
One could also argue that 3 is equivalent to 1 and 2, at least if it is
clear that observations are not used to shape the theory and/or that the
theorists didn't know about the observations.

Some confusion arises in debates because of failure to distinguish 1 and
2 (and perhaps 3) from 4 and 5.

An alternative assessment is that more confusion
arises from the attempt to pass off 2 and 3
as bona fide instances of 1.

No matter how you parse it and employ semantics,
2 can only generate retrodictions. In principle
there could be very strong retrodictions from 2,
but they cannot be true definitive predictions if
they are after the fact of observation.

Treating 3 as equivalent to 2 is a slippery slope.

Treating 3 as equivalent to 1 is abhorrent to science.

4 and 5 are model-building wherein you do not
understand the phenomena, but only hope to mimic it
in order to make some initial progress.

[Mod. note: reformatted. -- mjh]

In article , "Robert L.
Oldershaw" writes:

On Saturday, May 3, 2014 2:32:27 AM UTC-4, Phillip Helbig---undress to
reply wrote:

One could also argue that 3 is equivalent to 1 and 2, at least if it is
clear that observations are not used to shape the theory and/or that the
theorists didn't know about the observations.

Some confusion arises in debates because of failure to distinguish 1 and
2 (and perhaps 3) from 4 and 5.

An alternative assessment is that more confusion
arises from the attempt to pass off 2 and 3
as bona fide instances of 1.

I don't think anyone tries to "pass them off" as 1, but rather one
could claim that they are in some sense equivalent. Suppose a theorist
is working on the derivation of some quantity which can be measured
experimentally, and the experimentalists are working on the measurement
at the same time. Does it matter who gets there first? Does it matter
if the experimentalists know the prediction before they measure it?
Does it matter if the theorist knows the result before he derives it?
What if there is a vow not to disclose results to the other camp until
both are finished? Is it OK if the two camps don't know of each other's
existence?

Sure, a theory which has so many free parameters that it can fit
anything has no predictive power, and its retrodictive power is not
worth much either. But that's not what I'm talking about. I gave as an
example QED and the prediction and measurement of g-2.

No matter how you parse it and employ semantics,
2 can only generate retrodictions. In principle
there could be very strong retrodictions from 2,
but they cannot be true definitive predictions if
they are after the fact of observation.

Right, but no-one claims that they are. The only claim is that a
retrodiction which is so clear-cut that it could be a prediction, i.e.
there is no wiggle room, is just as good as a prediction.

On Monday, May 5, 2014 1:55:05 AM UTC-4, Phillip Helbig---undress to reply wrote:

I don't think anyone tries to "pass them off" as 1, but rather one

could claim that they are in some sense equivalent.
---------------------------------------------

If you read the literature carefully, you will find a
significant number of examples wherein a retrodiction
is referred to as a "prediction". This is true for
both the popular and technical journals, although the
most aggregious examples are found in the former.

Discussions of the BICEP2 results contained flagrant
examples.

"in some sense equivalent", I do not accept this
handwaving opinion, which I regard as a potential threat
to the time-honored scientific method.

One could regard a retrodiction as a valid prediction
if the theorist lived in isolation and had no
knowledge of the relevant observations. But how often
could this happen in our modern highly connected
world? Seems like a very far-fetched argument to me.

[Mod. note: not at all. It is very easy not to be able to find the
observation that tests a particular model in the vast literature
that's out there. Anyone who's written a number of papers has probably
had the experience of having relevant observations pointed out by the
referee or readers on arxiv. And I seem to recall some examples of
similar things on this very newsgroup -- mjh]

On Monday, May 5, 2014 1:55:05 AM UTC-4, Phillip Helbig---undress to reply wrote:

Right, but no-one claims that they are. The only claim is that a

retrodiction which is so clear-cut that it could be a prediction, i.e.

there is no wiggle room, is just as good as a prediction.
----------------------------------------

Here is my take on what has been slowly
taking place in fundamenal physics over
the last few decades. This is my personal
opinion and assessment.

There has been an increasing effort to
blur the distinction between true predictions
and retrodictions. A small but influential
group of celebrity physicists have promoted
the idea that the rules of science need to
change in our advanced era. The justification
is supposed to be that we are exploring realms
of very high energy, small/large space-times,
and the practically unobservable past. It is
argued that testability must be reconsidered
in light of trying to observe phenomena in these
semi-unobservable regions of space-time and
parameter space.

I think something very different is going on.
The standard models of particle physics and
cosmology are primarily model-building efforts.
They have served admirably for decades, but are
now showing the Kuhnian signs of reaching their
limits. The SM of particle physics has at least
seven serious shortcomings that make it clear
that it is a provisional theory. String theory
and supersymmetry have failed to come to the
rescue. The SM of cosmology is an effective
model of very general phenomena, but cannot
answer important questions without increasingly
weird and untestable hypotheses.

So the trend to blur the distinction between
predictions and retrodictions is due to the fact
that we no longer have adequate physical
principles to guide research and are relying
on model-building efforts that are increasingly
unable to generate definitive predictions.
When pseudo-predictions fail (sparticles) the
pseudo-predictions are just pushed into an
unobservable portion of the parameter space.

Proponents of the existing paradigm realize that
their models are increasingly untestable, so they
are beginning to invoke retrodictive and aesthetic
criteria in place of the standard scientific method.

I say we are approaching an era of Kuhnian
paradigm-shift, if we are not already deeply
within one.

In article , "Robert L.
Oldershaw" writes:

Here is my take on what has been slowly
taking place in fundamenal physics over
the last few decades. This is my personal
opinion and assessment.

Science advances when a rival theory comes along which explains
everything the standard theory does, but does so more elegantly and/or
makes testable predictions which a) differ from those of standard theory
and b) are then confirmed. It is the burden of those who criticize the
standard theory to come up with an alternative, and to accept defeat
when their alternative theory fails a prediction. One has to apply the
same standards to both theories. One cannot accuse one of introducing
epicycles but try to justify this for the other.

I say we are approaching an era of Kuhnian
paradigm-shift, if we are not already deeply
within one.

You speak of Kuhn as if it is generally accepted that his analysis is
correct.

On Thursday, May 8, 2014 4:09:41 AM UTC-4, Phillip Helbig---undress to reply wrote:
Science advances when a rival theory comes along which explains
everything the standard theory does, but does so more elegantly and/or
makes testable predictions which a) differ from those of standard theory
and b) are then confirmed. It is the burden of those who criticize the
standard theory to come up with an alternative, and to accept defeat
when their alternative theory fails a prediction. One has to apply the
same standards to both theories. One cannot accuse one of introducing
epicycles but try to justify this for the other.

One must bear in mind what is clearly
demonstrated in the historical record.

A small army of proponents whose status
is deeply linked to the prevailing paradigm,
make every effort to promote that paradigm,
and offer rationalizations for its shortcomings.

Moreover, they are highly motivated to
treat alternative paradigms as potential
threats, and to take every opportunity to
emphasize perceived shortcomings in the
new paradigms.

[Mod. note: I think it is time for this thread either to return to
astrophysics or to stop. Sociology of science can be discussed in
other fora -- mjh]

I strongly object to the premise of most of this discussion. In
science, the word "predict" has nothing to do with foretelling the
future. Instead it means deriving the consequences of a
theory. That's what we mean, for example, when we say Kepler's Laws
_predict_ Tycho Brahe's observations, which were done prior to
Kepler's derivation of his laws. In at least one way it's better
when the data come before the theory; the danger of "confirmation
bias" is thereby avoided.

The real question is not the time order of theory and observation but
rather how many adjustable parameters the theory has, how well or
badly its predictions match the data, and how extensive the data are.
In the Kepler case, there were 6 free parameters for each planet and
2 (if I've counted right) for the Earth (or Sun, if you prefer). With
those, Kepler was able to match hundreds or thousands of observations
of each planet to (in his final iteration) the observational accuracy
of about 2 arcmin. Newton gave us a theory with more free parameters
(the planets' masses), but the match with observations (when they
improved) was better. Actually it's not as bad as that -- some of
the planetary masses were determined independently from their
satellite orbits and nowadays all of them are known from spacecraft,
so they are not really free.

I see there's a new simulation that goes quite a way to predicting
cosmological observations:
http://www.cfa.harvard.edu/news/2014-10
I haven't read the article yet (and had nothing to do with the
research itself). If the work really has the galaxy morphologies and
chemical abundances right, that will be a huge achievement.

--
Help keep our newsgroup healthy; please don't feed the trolls.
Steve Willner Phone 617-495-7123
Cambridge, MA 02138 USA