What do do with the unphysical weight on the Omega-Lambda plot. When you have one parameter, you just suck it up that you've got unphysical probability (it seems). The problem is with two parameters; the unphysical part of Omega is giving weight to Lambda... is that what we really want? One suggestion is to use Omega_M=0 as the best _physical_ fit for all those points which fall at Omega_M<0.
Another thing in these plots, Peter looked at the minimum of the absoulte value of the residuals. They come in clumps, not in a smooth distribution (as is the case with looking at the square of residuals). By taking the power on chi up to 1.5, it starts to smooth out. No matter what power you look at, however, most of the probability falls at the same value, i.e. the whole distribution doesn't move.
Peter has not yet tried to do this fit with the Hamuy data in there as well.
Not only the unphysical data, but also the non-symmetric error bars; Peter thinks we need to talk to a statistician. There is also this issue of fitting in redshift space instead of flux (or magnitude) space, the astronomical community seems to be settling on the idea that you are more robust with respect to Malmquist bias.
Regardless of z vs. flux, we should do FLUX rather than MAGNIUTDE, since the errorbars are symmetric and gaussian in flux, but not in magnitude.
Next search, and issue of why we didn't do as well as we thought we would last run, and how we should approach the next run. (What is our strategy, do we want to do an I-band search instead of R-band, etc.)
Greg talked to Broadhurst; they didn't get any time. Gordon Squires, however, did get some time right before us. Greg will talk to him to find out if there is any way we can have any kind of overlap; we would then have the option of using those in March. We have to find out what he is going to do to find out if we can get anything out of this.
Some discussion of how to do fringe correction if we do do an I-band search. Can use a fringe map from the reference run... if the chips are the same! Some of us are a little nervous about doing our "last" search in I-band, i.e. moving to a new wavelength and all of the problems that could come with that when we have to find SNe for HST.
Two or three patches on the sky? Two patches=some high airmass imaging, three patches=things more divided up and a better chance of statistics zapping us in terms of where the SNe are and where we told HST they'd look.
Another arguement in favor of R: there is a feature in the U in SNe, Peter tells us, that gets mongo right before max. To take advantage of this in the I-band we'd have to move out to a redshift of around 1.
The other group didn't find that many SNe, due to weather losses at CTIO. However, they have 3 confirmed Ia's, two of which are z>0.8. They searched in I-band. It's not clear how much one could compare to this.
Saul is proposing we do half our search in R, to get half the SNe we did last time, and then do half the search in I to pound on higher redshift SNe. Or, perhaps, 1/3 in R at the same depth as last time, 1/3 in R at half-depth, and 1/3 in I.
Before we fully decide this, though, we have to figure out (1) what was going on with the depth of this run and (2) what we would have found scanning last spring's I-band data.
Re: depth, one question is electronic noise. This should show up as structure in the bias. Somebody should look into this.
Malmquist bias dicsussion: Greg asks, if we measure the stretch and correct, are we still hurt by Malmquist bias? The answer is yes, to the level of the post-stretch-correction intrinsic supernova dispersion, and also to our error in measuring the stretch. However, the inhomogenous nature of the universe (e.g. the big sheet at z=0.83) vs. the homogeneous nature of our volume sampling helps us.
Saul is more worried about the Malmquist bias in the Hamuy set. Determining that is harder. Saul suggests that we pretend our set is the real distribution of SNe and then find out how many of those Hamuy would detect. (This sounds pretty hard, actually.) That would measure the differential Malmquist bias... if you could do it right.
Take the peak apparent magnitude from our set, pretending we have a completely unbiased search, assume an Omega and Lambda, and move it back to the distribution of redshifts of the Hamuy set. How bright are they at peak at the Hamuy redshift? Compare this to their thresholds. See which ones we find and which ones we throw out. One issue is the kind of gap they have, so we know if we should use the discovery apparent magnitudes or the peak apparent magnitudes.
The goal is to decide if they are more or less sensitive than we are. If they are more sensitive then we are, then we move even further from the flat Lambda=0 universe. The suspicious is that since Hamuy tends to see lower stretches than we do, that they are not less sensitive than we are.
NOte of Stretch vs. Galaxy type. Peter states that in spirals you see all stretche but you are slightly more heavily weighted towards higher stretch SNe. In ellipticals you pretty much just see from the middle of the stretch distribution to thin stretches. Greg says looking at the Hamuy set, it looks like the dispersion of stretches is higher for ellipticals... which contradicts what Peter is saying./P>
Well, OK, looking at a 1996 Branch paper, it looks like the scatters are similar. Ellipticals have the larger delta_m15's.
This got pretty gory. The whole point of the thing is for those who don't like the stretch correction, but want to choose a set of SNe which are homogeneous based on host galaxy color, and just use those as standard candles rather than calibrated candles.Greg's plot of Stretch vs. host B-V. (Still have to iron out the R-I vs. B-V K-corrections, due to discrete colors.) In Hamuy, the redder galaxies have the narrower stretches. In our data, it's more flat; even in the redder galaxies we are seeing higher stretches. (Is this Malmquist?)
The other thing he had Robert working on was looking for evidence of more reddening in bluer galaxies; there evidence suggests that there is not. (This needs to be redone in E(B-V) rather than just B-V of the supernovae.) To Peter this suggests that we're severely Malmquist biased. To Saul this suggests that you have either no extinction, or you go through a whole magnitude of dust. Greg says this is consistent with the dust scale height in a galaxy being small in comparison to the star scale height.
The idea is to come up with a set which is of the scope of the Hamuy set, but without all the problems.
Greg plots up a table. He assumed a 15-day window so that we get them before max. He used Reynald's rates (0.8 SN/100yr/10^10 L_sun). He used the local luminosity distribution of the universe. He used a fiducial area of 50 square degrees. He went out to z=0.2. He used h=0.7. (Those last three give you a volume in space.) He got 15 Type Ia SNe (within a factor of 2). EROS is on the low end of this, but not entirely inconsistent.
The issue, then is how do you want to probe that much volume.
The followup time necessary is proportional to zmax^4 (ignoring things like moon and such). This (AND things like moon) would suggest that you want to do a shallower search.
Deep and narrow or wide and shallow? Which you do depends on which telecope you use. Has to do with how efficient you are, cycle time, exposure time, area of detect. Without readout time, it's always faster to go wide and shallow (i.e. if you only worry about photon noise). At z=0.25, peak magnitude would be 22, and we'd try to detect at 23. At mag 21, you're at z=0.16, at mag 20, you're at z=0.11 (or something).
He has plots which show how long you have to go and what your typical magnitude is. Brighter magnitudes mean easier followup.
Could even do a combination if you want to have a running search which is robust with respect to weather.