New WMAP Likelihood code ?

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Alessandro Melchiorri
Posts: 129
Joined: September 24 2004
Affiliation: University of Rome
Contact:

New WMAP Likelihood code ?

Post by Alessandro Melchiorri » October 23 2006

I just got this message from Antony Lewis as an update for CosmoMC:


"The new WMAP code is available at

http://lambda.gsfc.nasa.gov/product/map ... od_get.cfm

and no longer requires HEALPIX. CosmoMC changes are only trivial
changes to the Makefile and documentation - old versions should run fine
with the new code. sigma_8 is slightly higher with the new version."

Is the only place where I heard about this...what is the change in sigma_8 ?
What is the change in n_s ?

Cheers and thanks
Alessandro

Hans Kristian Eriksen
Posts: 60
Joined: September 25 2004
Affiliation: ITA, University of Oslo
Contact:

Re: New WMAP Likelihood code ?

Post by Hans Kristian Eriksen » October 23 2006

Alessandro Melchiorri wrote: Is the only place where I heard about this...what is the change in sigma_8 ?
What is the change in n_s ?
Well, it shouldn't come as a total surprise, I think.. :-)

There are two main changes, corresponding to the two issues described in [astro-ph/0606088] and [astro-ph/0606538]. Cosmological parameters for extended models were discussed in [astro-ph/0608017]. At low l's, there was a slight bias between l=13 and 30 when using the MASTER based approximation, while at high l's, the first WMAP3 release over-estimated the point source amplitude.

These two issues have now been corrected, although they use almost the mean of their original point source amplitude (A=0.017) and our best-fit amplitude (A=0.011) this time around. I'm looking forward to see how they derived this, because we would have preferred an even lower value -- the power spectrum between l=100 and 200 pulls the fit high, but this is likely to be non-point source effects. Also, the difference between the V and W-bands is bigger because of this higher value than what we found. Again, it will be interesting to read their justification for doing this, and I would be surprised if the referees will let this new value pass just like that.. :-)

However, I do agree that the publication policy this time was a little surprising. I expected the revised papers to appear on astro-ph before the likelihood code appeared on Lambda. Anyway, Jostein Kristiansen is working right now on getting the new parameters to see how they compare with the ones we found in our papers, and I'm sure he'll report the results here once he's ready.

Alessandro Melchiorri
Posts: 129
Joined: September 24 2004
Affiliation: University of Rome
Contact:

New WMAP Likelihood code ?

Post by Alessandro Melchiorri » October 23 2006

Thanks ! It will be great to have a more quantitative idea about the effect !
It is strange because usually there is an alert on the CMB section in cosmocoffee for any update on Lambda. Did I missed this news or it was'nt there this time ?

cheers
Alessandro

Jostein R. Kristiansen
Posts: 4
Joined: February 14 2006
Affiliation: University of Oslo
Contact:

New WMAP Likelihood code ?

Post by Jostein R. Kristiansen » October 23 2006

I've run cosmomc with a standard six-parameter model and the new WMAP likelihood code from LAMBDA. The most notable effects are on the amplitude and [tex]n_s[/tex]. I compare the results with the old WMAP likelihood code and the results from the point source and low-l corrected code used in astro-ph/0606538.

For n_s this gives:
old WMAP code: [tex]n_s = 0.954 \pm 0.016[/tex]
astro-ph/0606538-code: [tex]n_s = 0.966 \pm 0.016[/tex]
new WMAP code: [tex]n_s = 0.959 \pm 0.016[/tex]

For [tex]\log(10^{10}A_s)[/tex]:
old WMAP code: [tex]3.019 \pm 0.067[/tex]
astro-ph/0606538-code: [tex]3.039 \pm 0.068[/tex]
new WMAP code: [tex]3.029 \pm 0.068[/tex]

So it looks like the effect on the parameter estimation is significant, although smaller than what was found in astro-ph/0606538. So using the new WMAP likelihood code the significance of [tex]n_s \neq 1[/tex] is notable lower than what was found with their old code.

Jostein

Hans Kristian Eriksen
Posts: 60
Joined: September 25 2004
Affiliation: ITA, University of Oslo
Contact:

Re: New WMAP Likelihood code ?

Post by Hans Kristian Eriksen » October 23 2006

Jostein R. Kristiansen wrote: For n_s this gives:
old WMAP code: [tex]n_s = 0.954 \pm 0.016[/tex]
astro-ph/0606538-code: [tex]n_s = 0.966 \pm 0.016[/tex]
new WMAP code: [tex]n_s = 0.959 \pm 0.016[/tex]
OK.. I guess we'll have to wait for the revised papers and the point source amplitude justification in order to know what to think then.

But it is very interesting to look at those numbers like that -- the largest difference between any two estimates is 0.012, while the uncertainty is in both cases the same, namely 0.016. Makes me wonder how big the (unquantified) systematics uncertainty is.. :-)

Post Reply