From: Serena Nobili (serena@physto.se)
Date: Mon Feb 02 2004 - 05:24:07 PST
Dear Vitaliy,
thank you very much for your work. I think we are getting closer to the
final numbers for 1999Q. What I find annoying however, is the fact that we
don't get the same tmax as Tonry, given we are using the same data set.
From this point of view, Ariel and I have noticed that there are data
points in the lightcurve you are fitting which are above day 40 (or even
50). Those points are certainly worsening your chisq, since they are
always deviant. As the standard SCP procedure however, is not to include
points later than 40 days after max in the fit (I believe this was the
case also for Knop et al.), I think you should try to compute the chisq
value for all the cases excluding the latest 1 or 2 points.
Looking at the residuals in your plots, I expect this to give about the
same chisq in all the cases (except case 1 perhaps). If this is true, and
we get the same result as Tonry, than we can more easily explain in the
paper what we did. Thank you again for your help.
Cheers
Serena
On Fri, 30 Jan 2004, Vitaliy Fadeyev wrote:
>
>
>Hi All,
>
>first of all, Serena is right, I made a mistake in inferring the time
>of maximum from Riess's paper. Riess's JD is ...195.15 and Tonry's ...194.4,
>i.e. they are fairly similar.
>
>I went back and refit the lightcurve while fixing the day of maximum to the Tonry's
>value. This is case 3 below, the data contain the discovery point. You can see
>that the Chi^2/#(dof) is really bad. However, previously I did not consider
>the systematic effects in the photometry. Riess "... conservatively adopted
>a systematic uncertainty of 0.03 mag in the SN photometry...". I'm not sure
>what that actually means. Could imagine that a) this is a correlated error
>for all epics, or b) this is a "random" additional error. It did not seem that
>either case would change the conclusions. To check that this is true, I followed up
>with scenario (b). Cases 4 and 5 below repeat cases 2 and 3 with 0.03 mag
>added to the errors. I think the results with and without systematics
>are comparable.
>
>My conclusion is that the data prefer the earlier day of max that either Riess
>or Tonry reported. If day of max is ajusted to Tonry's value, then the stretch
>value is substantially lower (closer to 1).
>
>Cases are:
>1 - HST only
>2 - HST and discovery point
>3 - HST and discovery point, maximum fixed to Tonry
>4 - HST and discovery point, 0.03 mag added to the errors
>5 - HST and discovery point, 0.03 mag added to the errors, maximum fixed to Tonry
>
>SNMINUIT output table:
>case m_B dm_R m_B_corr dm_B_corr s ds chisq dof m_R kcorr_R kcorr_I tmax dtmax R-I d(R-I)
>1 22.82 0.06 23.02 0.16 1.117 0.054 5.29 8 22.06 -0.764 -0.310 2451186.54 1.444 -0.054 0.074
>2 23.14 0.02 23.34 0.07 1.119 0.032 9.95 9 22.24 -0.902 -0.419 2451191.83 0.853 -0.188 0.035
>3 23.22 0.02 23.31 0.06 1.040 0.022 26.86 10 22.23 -0.987 -0.485 2451194.65 0.000 -0.269 0.031
>4 23.09 0.03 23.31 0.08 1.131 0.035 7.00 9 22.23 -0.866 -0.390 2451190.94 1.467 -0.153 0.048
>5 23.23 0.03 23.35 0.07 1.061 0.025 16.00 10 22.26 -0.973 -0.475 2451194.65 0.000 -0.257 0.036
>
>The snminuit plots are here:
>(case 1) http://www-physics.lbl.gov/~fadeyev/snminuit_sn1999q_HSTonly.ps
>(case 2) http://www-physics.lbl.gov/~fadeyev/snminuit_sn1999q_HSTdisc.ps
>(case 3) http://www-physics.lbl.gov/~fadeyev/snminuit_sn1999q_fixtmax.ps
>(case 4) http://www-physics.lbl.gov/~fadeyev/snminuit_sn1999q_HSTdisc_syst.ps
>(case 5) http://www-physics.lbl.gov/~fadeyev/snminuit_sn1999q_fixtmax_syst.ps
>
>Cheers,
>vitaliy
>
>
>
>
>Serena Nobili wrote:
>
>> Dear Greg (cc Vitaliy & co.)
>>
>> I am glad you had the possibility to talk directly to Tonry. Reading your
>> e-mail I went back to Riess et al (2000) to check the date of maximum he
>> gives. Looking at table 2 of Riess'paper, I get the date of B maximum to
>> be 51195.15, i.e. only 1 day later than Tonry's value, and not 3 as you
>> say and, if I am not wrong, Vitaliy was assuming in case 2 of the fit. I
>> believe the problem comes from column 2 being rest frame, instead of
>> observed frame, this would give the 3 days difference.
>>
>> All of this is suggesting that the time of max given by Tonry and Riess
>> agree within 1 day. I think we should repeat the lightcurve fit using
>> Tonry's time of maximum, if we want to use the stretch factor, and refer
>> to the time of max as private communication (though, it would be nice to
>> be able to reproduce this fit, since we are using the same data set).
>> Once we have done this, we can safely include the stretch of 99Q in the
>> I-band paper. Thank you for your help in solving this problem.
>> Cheers
>>
>> Serena
>>
>> On Fri, 30 Jan 2004, Greg Aldering wrote:
>>
>> >
>> >Hi Vitaliy and Serena,
>> >
>> >I spoke with John Tonry today concerning 99Q. The photometry he has
>> >consists of the discovery point and those from the HST images. (This is
>> >basically the same as in Vitaliy's thrid fit.) The discovery point has I =
>> >22.3 +/- 0.05 and the fitted date of max is 2451194.4. So, Tonry's date of
>> >max is 3 days earlier than the one used by Adam, and 2.6 days later than
>> >your fit using the discovery point (albeit using a larger value for the
>> >uncertainty).
>> >
>> >Tonry said it is possible that Adam might have a "secret stash" of
>> >additional data points (I found this an odd thing for him to say, but
>> >apparently they aren't so good about sharing data within their
>> >collaboration), but this seems unlikely.
>> >
>> >So, I think we can safely use 2451194.4, quoting Tonry 2003 & private
>> >communication, for the time of maximum light. Using this modifed date
>> >of max isn't enough to alter the maximum restframe I-band brightness
>> >by more that a few percent. So, it isn't enough to explain why 99Q is
>> >fainter. However, it does mean that we know that stretch as well
>> >as the HZSST does.
>> >
>> >Looking more carefully, I see that in the uncorrected Hubble diagram 99Q
>> >is only about 2-sigma too faint using the *inner* error bars. In all
>> >cases 99Q is beter than 1.5-sigma using the outer error bars (i.e.,
>> >including intrinsic error).
>> >
>> >All of this suggests to me that we may as well show both the uncorrected
>> >and stretch-corrected results in the I-band paper. It will not be
>> >controversial, but make the paper clearer and more complete. (Right now
>> >one wonders about, or is easily confused by, the use of stretch correction
>> >for the local SNe in the first half of the paper and then skirting of the
>> >stretch correction in the latter half of the paper.)
>> >
>> >Cheers,
>> >
>> >Greg
>> >
>> >
>> >
>
>
------------------------------------------------------------------------ www.physto.se/~serena Tel +46 8 55378661
Give free food at: http://www.porloschicos.com/ http://www.thehungersite.com/
This archive was generated by hypermail 2.1.4 : Mon Feb 02 2004 - 05:24:19 PST