The third year WMAP ILC map

[Kate Land]

I am a little confused about the third-year ILC map, and the different versions available.

On the Lambda website, for 3rd-year ILC map there is the file:
'wmap_ilc_3yr_v2.fits'
They say a final step bias correction is applied, and this map is reliable on large scales ([tex]\ell\le10[/tex]).

However, with the WMAP Likelihood Software a very similar ILC map is provided, and the corresponding [tex]a_{\ell m}[/tex]'s. The relevant files are:
'bias_corrected_pass2_3yr_ilc.fits'
'alm_tt_fs_r9_ilc_nopixwin_3yr.dat'

Can anyone tell me what this latter map is exactly, and how it relates to the released ILC map above? They are very similar but the difference looks like a kind of patchy mask and a dipole.

Thanks

[Ben Gold]

I afraid my real answer is that I don't know exactly, and I'm having some computer trouble and can't look at things directly. But I can ask a couple leading questions:

1) Does the "dipole" bit look like figure 2 of Hinshaw et. al. (the 3yr temperature paper)? There may be a difference between the two maps in how the scan strategy is accounted for.

2) Does the "patchy mask" resemble the region definition mask? There are a number of slight changes that have happened with the ILC regions.

There's also apparently some small-scale structure left in the bias correction, so a change in how that was done would leave a different pattern. I would be surprised if the two maps gave significantly different C_l spectra, though. These should be minor differences.

[Hans Kristian Eriksen]

I afraid my real answer is that I don't know exactly, and I'm having some computer trouble and can't look at things directly. But I can ask a couple leading questions:

1) Does the "dipole" bit look like figure 2 of Hinshaw et. al. (the 3yr temperature paper)? There may be a difference between the two maps in how the scan strategy is accounted for.

2) Does the "patchy mask" resemble the region definition mask? There are a number of slight changes that have happened with the ILC regions.

There's also apparently some small-scale structure left in the bias correction, so a change in how that was done would leave a different pattern. I would be surprised if the two maps gave significantly different C_l spectra, though. These should be minor differences.

I have no idea what happened here, but I computed the full-sky spectra with anafast for both maps (with and without monopole and dipole removal), and the difference was very small, even without a sky cut. So I agree with Ben -- this probably doesn't matter at all for the C_l's. But of course, Kate is probably more interested in morphology, and then this is probably not quite as fun.. :-)

[Kate Land]

Thanks guys.

1) Does the "dipole" bit look like figure 2 of Hinshaw et. al. (the 3yr temperature paper)? There may be a difference between the two maps in how the scan strategy is accounted for.

No the dipole is more in the direction of the CMB dipole (due to our motion).

2) Does the "patchy mask" resemble the region definition mask? There are a number of slight changes that have happened with the ILC regions.

Yes, kind of. The difference looks like a Kp2 mask, but with some gaps in the galactic plane. Can also see the small circular dots in the ecliptic plane from masking mars, jupiter, etc. I'll try to get a picture online...

Both these maps claim to be third-year ILC maps, and therefore I assume they both use whatever new analysis, masks, and regions that there are. But something varied between the maps... the bias correction?

So I agree with Ben – this probably doesn't matter at all for the Cl's. But of course, Kate is probably more interested in morphology, and then this is probably not quite as fun.. :-)

Yeah - the difference isn't much. But I am looking at the Axis fo Evil stuff and we get different things for these two maps for the very low multipoles. Essentially I want to know which map is 'correct'(!) or better to use.

It is curious that I can't find any mention of what this second map is exactly... or why it isn't just the 3rd year ILC map that they released on Lambda. So the questions remains: Why do the Likelihood codes use a different third year ILC map than the one released on Lambda? and What is the difference due to?

Cheers, Kate

[Kate Land]

Picture is available here:
http://freddie.new.googlepages.com/plot

[Hans Kristian Eriksen]

Yeah - the difference isn't much. But I am looking at the Axis fo Evil stuff and we get different things for these two maps for the very low multipoles. Essentially I want to know which map is 'correct'(!) or better to use.

Sorry that I can't help you there, but if I had to guess, I'd say that the likelihood map is a version that was generated at an earlier stage (and possibly with the old ILC region definition) than the one at Lambda, and they just forgot to update it before the release. It doesn't look good, of course, but it probably doesn't matter for their stuff.

However, you can perhaps turn things around, and use the fact that you get different results for the two maps to say something about the stability of the "Axis of Evil" stuff? Or possibly lack of stability, if the difference is significant? This will of course be even more interesting when the WMAP team tells us which is correct, and how the "incorrect" map was generated.

[Ben Gold]

Picture is available here:
http://freddie.new.googlepages.com/plot

That's definitely the ILC region pattern there in the galactic plane.

What are the units? If those are [tex]\mu K[/tex], then the difference is far below the error, so the two maps are essentially equivalent. Even if it's [tex]mK[/tex], I think 0.02 mK is about the scale of the quoted error on the dipole, and with cosmic variance that's not going to get any better. So both maps (as far as the dipole goes) are still within one sigma of each other, and it's not clear which is "better" (since nobody really knows what the true 'cosmic' dipole is).

As for which ILC method is "better", I'll try and see if I can find out which map has the more recent method. But my guess is Hans is right: if you want the "official" ILC map it's best to stick with what's up on LAMBDA; the map included with the likelihood software is "guaranteed" to be good only for likelihood purposes. But if your results are that sensitive to the map used, you probably ought to be investigating their robustness in general.

[Kate Land]

What are the units? If those are μK, then the difference is far below the error, so the two maps are essentially equivalent. Even if it's mK, I think 0.02 mK is about the scale of the quoted error on the dipole, and with cosmic variance that's not going to get any better. So both maps (as far as the dipole goes) are still within one sigma of each other, and it's not clear which is "better" (since nobody really knows what the true 'cosmic' dipole is).

The units are mK. And it does indeed look alot like the ILC regions map. Curious that there is also the mars, jupiter, etc... dots.

But if your results are that sensitive to the map used, you probably ought to be investigating their robustness in general.

Fair point. That's what I was thinking...

Thanks, Kate

[Kate Land]

I have recieved confirmation from a WMAP member that the 'bias_corrected_pass2_3yr_ilc.fits' map that came with the Likelihood code is an earlier version. The 'wmap_ilc_3yr_v2.fits' map on the Lambda website is the official up-to-date one.

A re-release of the Likelihood code should correct this, and a few other things.