Open drphilmarshall opened 7 years ago
Thank you so much!
For the second bullet point, I downloaded all the tables and realized that there was no information regarding velocity dispersions. In calculating synthetic magnitudes, the velocity dispersions are really important; could I just set all the velocity dispersions equal to 200km/s?
Ah: these might be in a different paper. Try Paper IV or V of the series, searching ADS for Sonnenfeld and Marshall.
On Fri, Jan 27, 2017 at 12:11 AM, Jenny Kim notifications@github.com wrote:
Thank you so much!
For the second bullet point, I downloaded all the tables and realized that there was no information regarding velocity dispersions. In calculating synthetic magnitudes, the velocity dispersions are really important; could I just set all the velocity dispersions equal to 200km/s?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jennykim1016/OM10/issues/3#issuecomment-275610663, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY9zeR0L4I1qBaarPn84nqvvZCbIvSks5rWaalgaJpZM4Lu9Nh .
Thanks! Paper IV contained the exact information about the velocity dispersion that I needed. I will merge the tables into one and draw the cornerplots.
Hi @drphilmarshall ,
I finished reweighting a data, and the results look even more reasonable in the middle. The blue(mock, colored quasars) got much closer to the red, real data. However, the very top plots still look a little weird, which I think it is mainly because of the size of the samples.
Would just manually checking the weighting algorithm be the best way to solve the problem? Currently, the algorithm seems fine, but I can investigate more tomorrow with a clearer brain.
Thank you so much.
That does look better! Are you doing the distribution-matching in redshift? I would expect the 1D distributions to match in both centroid and width in this parameter. (I note that you still need to do the re-weighting of the red histograms to make them have the same normalization as the blue)
On Mon, Jan 30, 2017 at 10:50 PM, Jenny Kim notifications@github.com wrote:
Hi @drphilmarshall https://github.com/drphilmarshall ,
I finished reweighting a data, and the results look even more reasonable in the middle. The blue(mock, colored quasars) got much closer to the red, real data. However, the very top plots still look a little weird, which I think it is mainly because of the size of the samples.
Would just manually checking the weighting algorithm be the best way to solve the problem? Currently, the algorithm seems fine, but I can investigate more tomorrow with a clearer brain.
[image: after_reweight] https://cloud.githubusercontent.com/assets/18297066/22454834/98cedfc4-e73d-11e6-97e6-d93334eeee6c.png
Thank you so much.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jennykim1016/OM10/issues/3#issuecomment-276288250, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY9wG-hVrK0Ya5ib7OpXdf8yQZv9gzks5rXtmrgaJpZM4Lu9Nh .
Dear @drphilmarshall ,
I normalized the red and blue histograms, and the four plots look much better, but the first one does not seem it is normalized. Would this be an acceptable plots?
I am not sure what the "distribution-matching" is, but I did weight the OM10 lenses to have gaussian distribution as its parent population distribution.
Thanks!
OK cool: in that case, just make sure the notebook has enough math (you can use latex delimited by $$ in the markdown cells) and explanatory text to describe what you did, and we'll digest it tomorrow. Nice work!
PS. I'm not sure why the first histogram is coming out oddly: maybe expand the axis range in the $i$ dimension to see if we're missing a lot of objects?
On Wed, Feb 1, 2017 at 3:09 PM, Jenny Kim notifications@github.com wrote:
Dear @drphilmarshall https://github.com/drphilmarshall ,
I normalized the red and blue histograms, and the four plots look much better, but the first one does not seem it is normalized. Would this be an acceptable plots? [image: normalized histogram] https://cloud.githubusercontent.com/assets/18297066/22530373/115ac018-e890-11e6-82f6-5cf04c313ab9.png
I am not sure what the "distribution-matching" is, but I did weight the OM10 lenses to have gaussian distribution as its parent population distribution.
Thanks!
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jennykim1016/OM10/issues/3#issuecomment-276813222, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY9x6yk3dFmhixtToUfvLNCnzWCzmNks5rYRCygaJpZM4Lu9Nh .
Yes, I will update the notebook accordingly. Thank you so much!
Hi Jenny!
Your lens color comparison is looking really good! I think the OM10 colors are not far from reality at all, looking at the z=0.3 region. I have two suggestions:
You could re-weight the OM10 lenses so that they appear to be drawn from a Gaussian distribution that has the same mean and stdev as the actual SDSS LRGs. This should enable a closer color comparison.
You could check the colors at higher lens redshift by comparing to the CFGTLS colors given in Sonnenfeld et al 2013, SL2S Paper III. Again, you'll need to match the OM10 lenses redshift distribution in the same way as above.
You could do this by writing a
gaussian_reweight(parameter, (mean, stdev))
method indb.py
.