Evidence for w=-1 ?

Post Reply
Pablo Fosalba
Posts: 29
Joined: September 25 2004
Affiliation: Institut d'Estudis Espacials de Catalunya
Contact:

Evidence for w=-1 ?

Post by Pablo Fosalba » October 07 2005

Hi,

I just run some MCMC chains with the usual 6D paramater space + w,
and used WMAPext+SDSS data. I was surprised to find that the best fit
value for w is w=-0.3 +-0.13 (symmetric errors). I used no prior on HST or tau. When I cross-checked with the literature, e.g,Tegmark et al. astro-ph/0310723 I could not verify this result since they use the SN data too to constrain w and get w=-1 with small errors (see their Table 3).
So, has anybody done this excercise before?
this seems to suggest that the evidence for w=-1 does not come from CMB+LSS data, but rather from SN data alone.

Alternatively, if one uses the HST prior, I think one "forces" the best fit
to w=-1, but that this is an external prior....

I have not checked whether WMAPext+2dF yields the same results,
but I wd be surprised if they didn't.

Pablo

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Evidence for w=-1 ?

Post by Antony Lewis » October 07 2005

We did something similar in the B03 paper (astro-ph/0507503), and results for SDSS+WMAP+w were indeed odd when you marginalize over the bias parameter. The odd large w models however have rather rediculous values of the bias (not near one), so if you impose a crude bias prior - as we did in the B03 paper - the result becomes more sensible. (the CosmoMC default is to marginalize over b^2 with an infinite flat prior)

2dF+WMAP I think gives much more sensible looking results. But why this difference between 2dF and SDSS??

Jochen Weller
Posts: 45
Joined: September 24 2004
Affiliation: Ludwig-Maximilians-University Munich
Contact:

Evidence for w=-1 ?

Post by Jochen Weller » October 08 2005

You might want to have a look at the Moss and Battye paper:
astro-ph/0503033

Pablo Fosalba
Posts: 29
Joined: September 25 2004
Affiliation: Institut d'Estudis Espacials de Catalunya
Contact:

Evidence for w=-1 ?

Post by Pablo Fosalba » October 10 2005

Good point Jochen.
Actually, If one replaces SDSS by 2dF (and no SN Ia data), the best-fit value is w=-0.8
This suggests that there is some tension between SDSS and 2dF as far as w is concerned. This tension goes away once one assumes a prior on the bias as pointed out by Antony.
In any case, its unclear to me what is the fundamental difference between SDSS and 2dF...

Pablo

Niayesh Afshordi
Posts: 49
Joined: December 17 2004
Affiliation: Perimeter Institute/ University of Waterloo
Contact:

Evidence for w=-1 ?

Post by Niayesh Afshordi » October 13 2005

It is easy to see analytically that for w=-1/3 and c_s=0, the ISW and SW effects exactly cancel each other for long wavelength modes. My guess is the this, along with low WMAP quadrupole/octopole drives the the values of w=-1/3 and small c_s in CMB only analyses, such as that of Battye & Moss (astro-ph/0503033).

The curious thing is of course that SDSS doesn't seem to contradict this (but 2dF does!!).

Now, does anybody have any idea if the baryonic wiggle detections (in SDSS and 2dF) chnage this result?

Michael Doran
Posts: 41
Joined: November 22 2004
Affiliation: ITP Heidelberg
Contact:

Re: Evidence for w=-1 ?

Post by Michael Doran » October 18 2005

Pablo Fosalba wrote: I just run some MCMC chains with the usual 6D paramater space + w,
and used WMAPext+SDSS data. I was surprised to find that the best fit
value for w is w=-0.3 +-0.13 (symmetric errors).
I've done such "excercise" quite frequently and my bottom line is that CMB+LSS are not very sensitive to w, i.e. in particular w = -0.3 +-0.13 seems too tight: they "like" values closer to w=-1 but more of the w = -1 + 0.4 kind.

The w < -1 come entirely from Sne Ia, nothing else.

But now I have a question: Do you leave a free bias between CMB and LSS, or what kind of bias do you use ? And which SDSS and 2dF data points do you take into account ? When do you cut, because of non-linearities etc. ?

Michael

Pablo Fosalba
Posts: 29
Joined: September 25 2004
Affiliation: Institut d'Estudis Espacials de Catalunya
Contact:

Evidence for w=-1 ?

Post by Pablo Fosalba » October 20 2005

I've done such "excercise" quite frequently and my bottom line is that CMB+LSS are not very sensitive to w, i.e. in particular w = -0.3 +-0.13 seems too tight: they "like" values closer to w=-1 but more of the w = -1 + 0.4 kind.

The w < -1 come entirely from Sne Ia, nothing else.

But now I have a question: Do you leave a free bias between CMB and LSS, or what kind of bias do you use ? And which SDSS and 2dF data points do you take into account ? When do you cut, because of non-linearities etc. ?

Michael
Hi Michael,

The constrain you mention, w=-1+-0.4, is what you get with WMAP alone.
If you add 2dF data you should get something like w=-0.3 as I said.
Btw, I am using cosmomc datasets, so I use no prior on the galaxy bias and the scales included in the analysis are linear I believe.

I have another question:
Does the SN Ia data have an implicit HST prior on h ?
its seems all parameter constraints including SN Ia are consistent with
such prior...

Pablo

Michael Doran
Posts: 41
Joined: November 22 2004
Affiliation: ITP Heidelberg
Contact:

Evidence for w=-1 ?

Post by Michael Doran » October 20 2005

Hi,

Sne Ia constrains do not depend at all on H: the distribution is flat, as it should be. So of course, Sne Ia are "consistent" with H=0.72, cause they are consistent with any H (as long as H is treated as nuisance parameter).

As far as 2dF is concerned: I used 2dF until about a year ago, then switched to SDSS for LSS constraints. I never found w=-0.3 +- 0.1 as a prefered value, if I recall correctly. If I ever find time, I'll start a CMBEASY chain with WMAP + 2dF only to cross check.

Pablo Fosalba
Posts: 29
Joined: September 25 2004
Affiliation: Institut d'Estudis Espacials de Catalunya
Contact:

Re: Evidence for w=-1 ?

Post by Pablo Fosalba » October 20 2005

Michael Doran wrote:Hi,

Sne Ia constrains do not depend at all on H: the distribution is flat, as it should be. So of course, Sne Ia are "consistent" with H=0.72, cause they are consistent with any H (as long as H is treated as nuisance parameter).

As far as 2dF is concerned: I used 2dF until about a year ago, then switched to SDSS for LSS constraints. I never found w=-0.3 +- 0.1 as a prefered value, if I recall correctly. If I ever find time, I'll start a CMBEASY chain with WMAP + 2dF only to cross check.
Hi,

thanks for the piece of info.
Actually I realized that, in my last message, where I said 2dF, I really meant SDSS. So, for CMB+SDSS I do get w=-0.3, whereas for CMB+2dF I obtain w=-0.8 as best-fit.
So, it would be definetely interesting if you could cross-check that.

Pablo Fosalba
Posts: 29
Joined: September 25 2004
Affiliation: Institut d'Estudis Espacials de Catalunya
Contact:

Evidence for w=-1 ?

Post by Pablo Fosalba » November 18 2005

I think I found out the origin of the tension between SSDS and 2dF data.
Basically, they measure slightly different values of the slope of the galaxy power spectrum what propagates onto different best-fit \Omega_m values. SDSS prefers a value close to 0.3 while 2dF suggests significantly lower matter densities, 0.2
That's why the margnalised values for w from the banana-shaped contours for w-\Omega_m yield discrepant values of w too.

Actually, something similar to this was already discussed in a 2dF paper
by A. Sanchez et al. http://arxiv.org/abs/astro-ph/0507583

Pablo

Post Reply