Fig 2 of SEB paper for two cases Shen08 and Shen10:

here is the Fig 2 using Shen 2008 (Just redoing the Fig 2 in SEB paper for reference)

and then using Shen 2010 and exactly the same format as Fig 2:

well the SEB is NOT completely gone (the orientation of the distribution has not changed) but since Lbol Shen 2010 is scattered with respect to Shen 2008 so the M-L plane-scatter is larger this time.

There are little more objects here in this sample too, since they are not yet crosslisted with R&H sample but that doesn't change anything significantly.

Using sigma_line creates a counter-clockwise rotation with respect to Fig 2(new/old) which looks different in nature.

Shen et al. 2010 DR7 mass estimates vs 2008 DR3

at last I could read the fits table in Matlab. It seems that it works well.

Next step is read the 2008 data and cross-list them with 2010. here is a plot of log BH masses from Shen 2008 (76990 objects) crosslisted with Shen 2010

I was looking for a possible clockwise rotation in the above plot. If there was a clockwise rotation that could be interpreted that the new mass estimates using two Gaussian to fit the emission lines has resolved the difference between sigma_line and fwhm.

There is no such rotation in this plot that means using one Gaussian or two Gaussian doesn't change anything. The main source of the difference is non-Gaussianity of the lines.

Just to be sure that I am comparing same objects from both data sample I have plotted some other quantities like redshift, Lbol and at last MJD,PLate,Fiber. So there is no problem with the crosslisting.

 

Surprisingly their measured Lbol is different now.

 

 

 

TO DO NEXT:

about catalogue paper, I need a to have an object with low S/N ratio to see the effect of the PCA. which means before PCA we estimate larger sigma but after PCA sigma is smaller.

SMBHs do not correlate with galaxy disk or psuedobulges by Kormendy et al. 2011

Jesse, Laura and I had the quasar club. the minutes are written in Jesse's blog.

http://jesserogersonresearch.blogspot.com/2011/01/minutes-quasar-club-kormendy-et-al-2011.html

SEB Paper: referees point:

Peterson has asked "There is, however, one important point where I remain confused (suggesting that other readers might be similarly confused) and perhaps the authors can clear this up. It is not perfectly clear to me where the breakdown in the Shen et al masses actually occurs: It is not clear whether the major contributing factor is (1) use of FWHM rather than line dispersion, (2) the assumption that the line profiles are reasonably characterized by a Gaussian fit, or (3) the combination of the two."

okay, I guess I know how to investigate this. We have Shen et al. 2004 using one Gaussian fit and Shen et al. 2010 using two or more Gaussian fit. Comparing these two will clear situation.

1) if they are mostly the same then using FWHM will be the main source of difference

2) if they are not the same and indeed the new estimates are more close to what line dispersion gives then the main source is the number of Gaussian approximation.

Now I am going to investigate this.

218 spectra:

dereddening is applied on 218 objects.

The Program is adjusted to use dereddend spectra now and is working well so far.

Dereddend DR7 spectra:

Pat has applied the dereddening on DR7 except the 218 objects added by me.

I have changed the permissions of the objects in plates:

2352, 0733,0734,0736,0737,0739

Here is what Pat has done. I will follow him:

" the best way to proceed is to rerun the reconstruction
on dereddened spectra.  Now, earlier I mentioned that the extinction curve
that's probably best is that of Fitzpatrick, E. L., 1999, PASP 111, 63.
But Yip et al used O'Donnell and CCM; I don't know the O'Donnell reference
but CCM is Cardelli,  Clayton,  and  Mathis,  ApJ  345:245,  1989.

The difference between extinction laws is small in the optical, and the
extinction corrections are small at the high Galactic latitudes of the SDSS.
Small differences to small corrections are not a big deal, so I think the
best thing to do about dereddening is the quickest thing.  That's to use
the IRAF task "deredden" which has the CCM law built-in.  I've run it on
all the spectra on canis that I can, so there are now dereddened versions
of almost all the spectra ready to go, including the duplicate spectra.
(It turns out that all the duplicate spectra of DR7 quasars were already
on canis.)

The only objects which I wasn't able to deredden are the 218 objects whose
spectra you added to canis, because only you have permission to read those
spectra.  To deredden those objects yourself, start IRAF, type "noao" and
"onedspec" to load packages, cd /data/phall/WILD/SKYSUB/ and type:
        cl < finishDR7dered.cl
...and then change the permissions on the original and deredenned spectra
so that everyone has read and execute permission on them.

The dereddened spectra are in the same directories, with "DR" appended
to their names.  That is, you should reconstruct
2152/spSpec-53874-2152-162_skysubDR.fit
instead of
2152/spSpec-53874-2152-162_skysub.fit"

NATS1740 marks correction day.

There was around 20 students with some sort of problem in their marks. I sat in my office to correct their marks or re-evaluate the assignments. It was a busy day. I have Jan 24th for some more students to come and this course will be over after that.

In between, I had time to look on some of the reconstructed spectra. I can definitely tell that the eigenspectra are already corrected for the dereddening.

I will take care of the duplicated spectra soon. may be, I'll wait till I know what to do with dereddening. Now, it will take 10 days to repeat the reconstruction for all objects (we have all the objects and program doesn't need me to check it). so if I have to dereddend the spectra before reconstructing them, I think it is okay, I will repeat them now.

-----

This is what Yip et al. have applied:

assume standard Rv=3.1

default: O'Donnell's extinction curve in range [3030,9091] angstrom
          outside this range, use CCM (1989)

The foreground dust reddening was removed from the QSO spectra using
Schlegel et al. Milky Way dust map. We did not save the dereddened
spectra. The intrinsic reddening was not treated, which in turn would
show up in the PCA mode(s).

---

DR7 quasar spectra reconstructed

 

at last all 105,785 spectra are reconstructed.

Missing more objects:

plate 0736 and 0737 are empty too. Now I am wondering if there is something important behind this missing objects!

okay. I have also added 40 objects to Canis 0737 directory too.

plate 0739 is empty too. there are 43 objects in the list. These are also added to the Canis now.

I hope I am not destroying Pat's data on Canis by adding these data to them. 

New problems with the sample:

apparently there are some more objects missing from plate 0733

so far I have added few more objects to canis but program keep stopping now.

the question is, why there so much more objects in the plate 0733 online directory but less in Canis directory 0733? are they excluded from the directory for some reason or they just left out?

okay. I have added all 50 objects which was missing from the list in to the Canis directory.

there are 44 objects missing from plate 0734!

okay. they are added to Canis now.

PCA is working now:

Started the program on Friday night.

it stopped for unknown reason on Monday night on object 39578 with SDSS_ID and pate/mjd/fiber  (in catalogue list  pbh_dr7ANDhw10_dr7qzHW)

103453.55+241332.8     2352/53770/217

doing a little investigation: apparently there is no such file

spSpec-53770-2352-217_skysub.fit

in the /plate  directory! or any where else in SKYSUB directory.

continuing after that object (illuminated from the list for now!)

going back to DR3 mass-catalogue for now

PCA program is ready to run:

 

I use "pbh_dr7qANDhw10_dr7qzHW" catalogue. I read M_i and redshift (HW) and plate mjd and fiber and SDSS_J2000_ID from this file.

the program min3_qev.sm is adjusted to record  "Mibin" and "zedbin" as well as all 50 eigenspectra in order from 0 to 40 in vector "qsoEigCo" and the reduced-chi2.

I also record the reconstructed file. I will use this to calculate the MgII-mass later.

Problems I need to think about:

1) there are about 134,824 objects in DR7 directory in Canis; however, there are only less than 100,000 in the list catalogue "x.dat090616". I need the M_i for the rest of the them.

so either I have to use the small sample or if I have to use the big sample then I need to calculate the M_i the same way they have calculated for DR7 catalogue.

Pat says: leave the duplicted spectra for now. so we use pbh_dr7ANDhw10_dr7qzHW file from http://ara.phys.yorku.ca

Work on Catalogue Paper:

 

The figure 3 was suggested to be changed by referee so I have changed the fig3.a (top) color from cyan to black and the middle one to blue and cut the reference line zero from fig3.c (down)

(I use program in line  434 of "ref_fit_calibration.m"  located at  C:\Users\alireza\Work\Work In Progress\Research\P18 Re_fit_calibration. I used "AftbefPCA_sp52355_0574_016.eps" as fig 3)

I have answered one of the referees point as:

referee: >Similarly, since another component of this work is the advocacy of
>using the line dispersion over the FWHM, it would also be interesting
>to see if sigma_line(MgII) tracks sigma_line(Hbeta) better than
>FWHM(MgII) tracks FWHM(Hb) for the RM sample.

(Pat) TODO: make figure for inclusion and/or description in paper (these plots didn't help our argument)

(Ali): I should add this to answer to referee:
We have submitted another paper to MNRAS to address this point (the advocacy of using the line dispersion over the FWHM) and to distinguish potential biases caused due using FWHM(MgII). Since the subject is broader than just one paragraph or a section, we prefer not to adress it here.

------------------

these are remin for tomorrow:

TODO:
Figure out how many of 126 partial-duplication quasars should be
added back to the catalog and add them back with correct DR3c flags.
No need to mention the issue to the referee.

 

 

TODO:
[As for demonstrating, you could plot a few objects where a noise spike
was fit in the bb case but the full line was recovered in the ab case.]

(Ali) I can add this in addition to what I have in the top of this page.

 

------------------

Adjusting the PCA output:

 

I use min_qev.sm which in tern it uses the qev.sm which tern it uses qfiter for N=50 the number of eigenspectra.

inside "min_qev.sm"

1) the path_file address should be adjusted to the new address for eigenspectra source in Canis. (done)

2) the redshift and absolute magnitude in i band is required here to fill into qfiter program (I got the DR7 list includes the redshift and iMag; done)

3) out_name is created here too (done)

inside "qev.sm"

1) if input is fits format then we use qfiter(I prefer this)

2) if input is ascii file then we use qfiterT (it is in qev2.sm)

---------

what I have done:

1) I have a copy of the ES (the eigenspectra files) in canis now

/data/arafiee/ES

2) I have adjusted the qev.sm and min1_qev.sm on canis at /data/home/arafiee/work/DR7EVR

to record the coefficient (qsoEigCo) as well as the variance of the coefficient (qsoEigVar) and teh reduced chi squared (outchinu2) in a file with name qsoEigCo_MJD_Plate_Fiber

the reconstructed spectra is recorded in a file   ######qsoEigRecon_MJD_Plate_Fiber

3) the ES directory contains information for eigenspectra up to 50 eigenspectra. I need it up to 100 (since we want to post the coefficient it is better  if we go to higher number)

---------