A Space & astronomy forum. SpaceBanter.com

Go Back   Home » SpaceBanter.com forum » Astronomy and Astrophysics » Research
Site Map Home Authors List Search Today's Posts Mark Forums Read Web Partners

Speaking of Statistical Significance!



 
 
Thread Tools Display Modes
  #1  
Old June 5th 15, 08:08 PM posted to sci.astro.research
Robert L. Oldershaw
external usenet poster
 
Posts: 617
Default Speaking of Statistical Significance!

Check out this new submission to arxiv.org.

http://arxiv.org/abs/1506.01354

"Marginal evidence for cosmological acceleration from Type Ia supernovae"

The paper calculates that the significance for cosmological acceleration based on post-1990s SN Ia data is less than 3-sigma.

Counter-arguments?
  #2  
Old June 6th 15, 11:29 AM posted to sci.astro.research
jacobnavia
external usenet poster
 
Posts: 105
Default Speaking of Statistical Significance!

Le 05/06/2015 21:08, Robert L. Oldershaw a écrit :
Check out this new submission to arxiv.org.

http://arxiv.org/abs/1506.01354

"Marginal evidence for cosmological acceleration from Type Ia supernovae"

The paper calculates that the significance for cosmological acceleration based on post-1990s SN Ia data is less than 3-sigma.

Counter-arguments?


The first author of that paper (Mr Nielsen) specializes in data
analysis. Other papers by him in Arxiv show him as a very mathematically
minded astronomer. In another paper he develops a math system for better
images for the SAMI survey, in 2011 he publishes
"PyWiFeS: A Rapid Data Reduction Pipeline for the Wide Field
Spectrograph (WiFeS)", he worked in the CANDELS survey too.

The article about the supernova is VERY difficult to follow, but what I
understood means he uses new data about supernovae to put the
supposition that "the universe"'s expansion is accelerating into a
statistically more solid basis. And then he realizes that the
"acceleration" disappears!

What is interesting too in that paper is to see how sweeping
generalizations dominate cosmology. After seeing a few birds,
cosmologists start screaming immediately that the spring has arrived...

When cosmologists speak about "the universe", it would be better that
they have a reasonable sample size to backup their claims!
  #3  
Old June 6th 15, 11:31 AM posted to sci.astro.research
Phillip Helbig (undress to reply)[_2_]
external usenet poster
 
Posts: 273
Default Speaking of Statistical Significance!

In article , "Robert L.
Oldershaw" writes:

Check out this new submission to arxiv.org.

http://arxiv.org/abs/1506.01354

"Marginal evidence for cosmological acceleration from Type Ia supernovae"

The paper calculates that the significance for cosmological
acceleration based on post-1990s SN Ia data is less than 3-sigma.

Counter-arguments?


False dichotomy. You seem to be assuming that either this new paper is
completely correct or the old conclusions were completely correct, i.e.
if one believes the latter then one must find some argument to show why
the former is wrong. It's not so clear-cut.

First, looking at FIG. 2 in the new paper, an accelerated universe is
still MUCH more likely than a non-accelerated universe. Their claim is
that, when all uncertainties are taken into account, that one can't
strongly rule out a non-accelerating universe. OK, but, again, even
their new analysis has an accelerating universe much more likely than a
non-accelerating universe.

However, a non-accelerating universe compatible with the supernova data
would have a very low Omega and a very low lambda (which could even be
negative, though they don't explore this). However, there is a huge
amount of evidence, COMPLETELY INDEPENDENT of any supernova data, which
points to the concordance model with Omega=0.3 and lambda=0.7.
According to this new paper, this is still a very good fit to the
supernova data, much better than any non-accelerating universe.

A paper like this would be really interesting only if it presented data
which ruled out the concordance model. Even then, the conclusion would
probably be that something is wrong with the supernova data, because the
alternative would be that several other independent cosmological tests
are all wrong and nevertheless agree.

Summary: They show that the uncertainties in the supernova-data analysis
might have been overestimated, but even so the concordance model is
still a good fit and acceleration is much more likely than not.
  #4  
Old June 6th 15, 11:31 AM posted to sci.astro.research
Phillip Helbig (undress to reply)[_2_]
external usenet poster
 
Posts: 273
Default Speaking of Statistical Significance!

In article , "Robert L.
Oldershaw" writes:

Check out this new submission to arxiv.org.

http://arxiv.org/abs/1506.01354

"Marginal evidence for cosmological acceleration from Type Ia supernovae"

The paper calculates that the significance for cosmological
acceleration based on post-1990s SN Ia data is less than 3-sigma.


From a completely different point of view, my latest paper
(arXiv:1505.02917) also argues that uncertainties in the cosmological
parameters are larger than is generally assumed.
  #5  
Old June 6th 15, 07:26 PM posted to sci.astro.research
Phillip Helbig (undress to reply)[_2_]
external usenet poster
 
Posts: 273
Default Speaking of Statistical Significance!

In article , jacobnavia
writes:

The article about the supernova is VERY difficult to follow, but what I
understood means he uses new data about supernovae to put the
supposition that "the universe"'s expansion is accelerating into a
statistically more solid basis. And then he realizes that the
"acceleration" disappears!


Wrong. Look at FIG. 2 in the paper. The difference between these and
most other papers on the topic is that a non-accelerating universe is
not ruled out at more than 3 sigma. However, even in this analysis, the
accelerating universe is is much more probable. FIG. 2 has the dotted
"no acceleration" line. BELOW that line there is deceleration, above it
there is acceleration. The paper can be summarized as "While previous
analyses ruled out non-acceleration at more than three sigma, this
analysis places weaker constraints, but an accelerating universe is
still strongly preferred by the data".

Nothing disappears!

What is interesting too in that paper is to see how sweeping
generalizations dominate cosmology. After seeing a few birds,
cosmologists start screaming immediately that the spring has arrived...


Wrong. The concordance model is called such because many lines of
evidence point to it. Even if the supernova data indicated strongly a
non-accelerated universe and ruled out an accelerated one, one would
still have to explain why so many other tests point to the concordance
model.

When cosmologists speak about "the universe", it would be better that
they have a reasonable sample size to backup their claims!


Again, wrong. Sure, when the sample size increases, better statements
can be made. However, if the error bars are calculated properly, then
smaller samples result in larger error bars, all else being equal. The
point of the paper is that the error bars have been underestimated in
the past.
  #6  
Old June 7th 15, 06:06 PM posted to sci.astro.research
jacobnavia
external usenet poster
 
Posts: 105
Default Speaking of Statistical Significance!

Le 06/06/2015 20:26, Phillip Helbig (undress to reply) a écrit :
Nothing disappears!


Excuse me but precisely the problem is that you were arguing not so long
ago that any result below 5 sigma was meaningless. Of course it doesn't
disappear, I wrote that too quickly. But below 3 sigma it is no longer a
relevant result isn't it?
  #7  
Old June 8th 15, 08:09 AM posted to sci.astro.research
Phillip Helbig (undress to reply)[_2_]
external usenet poster
 
Posts: 273
Default Speaking of Statistical Significance!

In article , jacobnavia
writes:

Le 06/06/2015 20:26, Phillip Helbig (undress to reply) a écrit :
Nothing disappears!


Excuse me but precisely the problem is that you were arguing not so long
ago that any result below 5 sigma was meaningless. Of course it doesn't
disappear, I wrote that too quickly. But below 3 sigma it is no longer a
relevant result isn't it?


One could argue that it isn't. Of course, a 3-sigma or 5-sigma cutoff
is arbitrary. Were there no other evidence for the concordance model,
then this result (assuming it stands up; I don't know if the paper has
been refereed yet nor where it was submitted) probably wouldn't convince
many people.

However, what this paper does NOT claim is that a non-accelerating
universe is favoured by the data. The data still clearly favour an
accelerating universe, though not as strongly as before, and a
non-accelerating universe is marginally allowed, whereas before it was
ruled out at well more than 3 sigma.
  #8  
Old June 8th 15, 04:57 PM posted to sci.astro.research
jacobnavia
external usenet poster
 
Posts: 105
Default Speaking of Statistical Significance!

Le 08/06/2015 09:09, Phillip Helbig (undress to reply) a écrit :
Were there no other evidence for the concordance model


I posted here on April 13th in the "Dark energy doesn't exist?" thread a
result that implied otherwise:

quote
The authors conclude that some of the reported acceleration of the
universe can be explained by color differences between the two groups of
supernovae, leaving less acceleration than initially reported.
end quote

There is mounting evidence AGAINST dark energy.

jacob
  #9  
Old June 10th 15, 07:22 AM posted to sci.astro.research
Phillip Helbig (undress to reply)[_2_]
external usenet poster
 
Posts: 273
Default Speaking of Statistical Significance!

In article , jacobnavia
writes:

Le 08/06/2015 09:09, Phillip Helbig (undress to reply) a écrit :
Were there no other evidence for the concordance model


I posted here on April 13th in the "Dark energy doesn't exist?" thread a
result that implied otherwise:

quote
The authors conclude that some of the reported acceleration of the
universe can be explained by color differences between the two groups of
supernovae, leaving less acceleration than initially reported.
end quote

There is mounting evidence AGAINST dark energy.


OK, let's assume the supernova stuff is wrong. Please explain why all
the other tests which indicate the concordance model a) are wrong and b)
just happen to agree with each other.

Again, there are a couple of papers claiming that the supernova error
bars might have been overestimated or whatever. Does anyone know the
status? Have they been peer reviewed? (If so, this does not mean they
are correct, but it means they have passed a hurdle of unknown height.)

It is one thing to say that the supernova error bars have been
overestimated. If so, then the change will be a larger allowed region
of parameter space, perhaps including non-accelerating universes,
negative cosmological constant, or whatever, but the best-fit value
probably won't change. It is something else to say that there is some
systematic source of error which distorts the results. If the latter,
then it is strange that one still gets a good fit with (some other)
cosmological model: with only two parameters to fit, and several hundred
data points, one might think that no conventional cosmological model
should give a good fit if something were really wrong.
  #10  
Old June 10th 15, 07:29 AM posted to sci.astro.research
Phillip Helbig (undress to reply)[_2_]
external usenet poster
 
Posts: 273
Default Speaking of Statistical Significance!

In article , jacobnavia
writes:

Le 08/06/2015 09:09, Phillip Helbig (undress to reply) a écrit :
Were there no other evidence for the concordance model


I posted here on April 13th in the "Dark energy doesn't exist?" thread a
result that implied otherwise:

quote
The authors conclude that some of the reported acceleration of the
universe can be explained by color differences between the two groups of
supernovae, leaving less acceleration than initially reported.
end quote

There is mounting evidence AGAINST dark energy.


Suppose things were reversed. Everything supports what you believe,
then ONE observation goes against it, at less than 3 sigma. Would you
believe it? Would you speak of "mounting evidence"? No. You would
probably say "one tentative result is not enough to convince me".

The situation now is that essentially all observations support the
concordance model, then one observation looks to be a bit less secure.
Yet you tout this as indicating "dark energy doesn't exist" or whatever.

The universe is what it is, regardless of what anyone believes it to be
or wants it to be. The only thing which is important is a proper
interpretation of observations. ALL observations.
 




Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
SETI Ignores Significance of Archaeoastronomy Mad Scientist Misc 3 August 31st 04 08:08 AM
SETI Ignores The Significance of Archaeoastronomy Mad Scientist Misc 7 August 27th 04 11:19 AM
Poll: Significance of WLE Upgrade Stuf4 Space Shuttle 2 April 5th 04 09:16 PM
Does 11 11 have significance astrometrics? Rob B Misc 4 March 13th 04 10:37 PM
11 11 Any significance astronomically speaking? Rob B UK Astronomy 1 March 13th 04 08:11 AM


All times are GMT +1. The time now is 04:57 PM.


Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 SpaceBanter.com.
The comments are property of their posters.