From John Greaves: === Brian As I quickly, but not detailedly, mentioned in the (way back) first mail on this thread, the SDSS r'_CMT mags _are_ on an AB system zeroed against Vega, from what I can gather from their preprint and guideline document. (So I suppose they ought to denoted SDSS r'_CMT_AB_(Vega), with subscripts upon subscripts). At least, I think that is what this cut and pasted quote is saying (some special characters and formatting will have been lost from the original .pdf): > Since the CMT only observes in one passband, photometry from the telescope > cannot be placed on a standard photometric system without additional colour > information. However, using the colour data in the Tycho 2 catalogue, it is > possible to calibrate onto the instrumental magnitude system. By using an a > priori colour term, the VT of the Tycho 2 standards can be converted into the > natural system of the CCD and filter combination. This is close to the Sloan r' > passband and will initially be on the Vega scale rather than the > spectrophotometric ABv magnitude system (Fukugita et al. 1996). However, > comparisons with the SDSS standards have shown no detectable colour terms. In > addition to this, the offset between the standards and the CMT magnitudes was > measured which enabled the photometry to be placed accurately onto the ABv > system. Figure 4 shows the comparison with the standards after the offset has > been applied, showing that the CMT r' magnitudes are well calibrated onto the > Sloan r' ABv system. So, from what you're saying, it seems surprising that the V fit seemed fairly okay. Any general thoughts on consequences of the above quoted details would be appreciated. GENERALLY on fitting tests / testing fits: Ondrej has kindly informed me that there is a quirk in the correlation coefficient re the goodness of fit testing in Excel, which is what I have been using. He suggests standard deviation would be better, which coincidentally I had started using, as it is more informative, as well as allowing some quantitative assessment of quality. The following forwarded bit explains the point better than I could. -----------start forwarded bit from Ondrej-------------------------------- This is because the slope of V fit is much higher than that of the Rc fit. I did a simpe Monte Carlo simulation. I created 10000 points to follow line with slope 0.2 and 0.7, respectively. I added Gaussian noise with sigma=0.15 to both datasets. Linear fit in Excel proves that despite the same scatter, dataset with lower slope has lower correlation coefficient (figures attached). [the figures gave R^2 ~ 0.9 for the steep slope and ~ 0.3 for the shallow slope JG] The simple standard deviation might be better estimate of "goodness of fit" though. ------------end forwarded bit----------------------------------------- F and J mags from GSC2.2 There's a bit on this on aavso list archives, I note. Basically, over and above all the valid reasons Arne and Brian give with respect to problems with this, the other main thing that has got to be remembered is that F and J mags in GSC2.2 from POSS II suffer from the self same problems as red and blue mags in USNO Ax.0. That is, they are not necessarily same epoch, and even if the star can be considered very constant, the conditions may well have been different at each exposure, and each exposure may well not have been taken on the same piece of kit. In equatorial zones they could be from schmidt palomar and/or UK schmidt and/or whatever the ESO people used in Chile. So why expect any connectivity. V from near contemporaneous same observatory USNO B1.0 B1 and R1 Northern Hemisphere POSS I plates has a standard deviation of +/- 0.5, _at least_, relative to V for objects in Brian's loneos.phot file, these particular mags givng the best possible case. Mixed blues and reds are worse case. Going to F and J isn't going to improve on that much. And this is over and above regional systematic problems that crop up from time to time. Cheers John
Return to the Powerful Daisaku
vsnet-adm@kusastro.kyoto-u.ac.jp