Comments on: Background Subtraction [EotW] http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/ Weaving together Astronomy+Statistics+Computer Science+Engineering+Intrumentation, far beyond the growing borders Fri, 01 Jun 2012 18:47:52 +0000 hourly 1 http://wordpress.org/?v=3.4 By: hlee http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/comment-page-1/#comment-257 hlee Tue, 24 Jun 2008 20:52:10 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=308#comment-257 I have a rough idea to handle background in a general way, yet I do not know how to realize my idea with actual background and source data. A system of equations should not force the assumption of background homogeneity unless there's a proof. I'm not opposing subtraction because to know the source flux, one should subtract background flux. Probably, I'll get a well known obs id (well defined background region and source of nonignorable background) from CDA (Chandra Data Archive) and start playing with it to check feasibility of my idea! Even though it turns out into a vain, it could serve as an indirect proof of homogeneous background. But I'm afraid to tell you that getting used to a new tool and learning science behind it takes sometime. I have a rough idea to handle background in a general way, yet I do not know how to realize my idea with actual background and source data. A system of equations should not force the assumption of background homogeneity unless there’s a proof. I’m not opposing subtraction because to know the source flux, one should subtract background flux. Probably, I’ll get a well known obs id (well defined background region and source of nonignorable background) from CDA (Chandra Data Archive) and start playing with it to check feasibility of my idea! Even though it turns out into a vain, it could serve as an indirect proof of homogeneous background. But I’m afraid to tell you that getting used to a new tool and learning science behind it takes sometime.

]]>
By: vlk http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/comment-page-1/#comment-251 vlk Wed, 11 Jun 2008 06:11:08 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=308#comment-251 r is deterministic, because it is the ratio of areas (in units of [pix2 cm2 sec]) in which the counts C and B are collected. There may be systematic uncertainties here, but no statistical uncertainties. b is not deterministic; it is the intensity of the background in the source area. Therefore, the area correction factor is unity, and there is no need to resort to extra variables t. You are correct that an assumption is made that the background at an off-source location also describes the background under the source. But if you do not make this assumption, you will have 3 variables (s and b in the source region and say b' in the background region) and 2 equations, so you will have to have another relation saying what b' is as a function of b. As long as b' is proportional to b, the proportionality constant can be subsumed into r and it will reduce without loss of generality to the same form as given above. If b' is a non-linear function of b, well, you shouldn't be using this formula then. r is deterministic, because it is the ratio of areas (in units of [pix2 cm2 sec]) in which the counts C and B are collected. There may be systematic uncertainties here, but no statistical uncertainties.

b is not deterministic; it is the intensity of the background in the source area. Therefore, the area correction factor is unity, and there is no need to resort to extra variables t.

You are correct that an assumption is made that the background at an off-source location also describes the background under the source. But if you do not make this assumption, you will have 3 variables (s and b in the source region and say b’ in the background region) and 2 equations, so you will have to have another relation saying what b’ is as a function of b. As long as b’ is proportional to b, the proportionality constant can be subsumed into r and it will reduce without loss of generality to the same form as given above. If b’ is a non-linear function of b, well, you shouldn’t be using this formula then.

]]>
By: hlee http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/comment-page-1/#comment-250 hlee Wed, 11 Jun 2008 03:54:26 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=308#comment-250 I wish I could address the problem with more sense, meaning modeling with statistically justifiable assumptions. The section 2.2 from this week's arxiv:astro-ph paper describes <strong>Background Subtraction</strong>, which I couldn't understand the procedure at the first glance (<a href=http://arxiv.org/abs/0806.1575" rel="nofollow">[astro-ph:0806.1575]</a>, more papers introducing background substarction are welcome). Is this the way how astronomers estimate b and s? My question is, what is random and what is fixed? C and B are random variables following the Poisson distribution with given rates, that I don't doubt. Yet, how come r and b are deterministic? How sure b is fixed and same for both equations in the first line? I bet checking S/N is to ensure gs is zero (or ignorable), though. All I wish to get is any historic account that b is homogeneous and two b's in the first line are the same (according to the account, r is a fixed value, independent of s, and I introduced t in tb to indicate that b is not necessarily homogeneous. The model would be written in different fashions). I wish I could address the problem with more sense, meaning modeling with statistically justifiable assumptions. The section 2.2 from this week’s arxiv:astro-ph paper describes Background Subtraction, which I couldn’t understand the procedure at the first glance ([astro-ph:0806.1575], more papers introducing background substarction are welcome). Is this the way how astronomers estimate b and s?

My question is, what is random and what is fixed? C and B are random variables following the Poisson distribution with given rates, that I don’t doubt. Yet, how come r and b are deterministic? How sure b is fixed and same for both equations in the first line? I bet checking S/N is to ensure gs is zero (or ignorable), though.

All I wish to get is any historic account that b is homogeneous and two b’s in the first line are the same (according to the account, r is a fixed value, independent of s, and I introduced t in tb to indicate that b is not necessarily homogeneous. The model would be written in different fashions).

]]>
By: vlk http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/comment-page-1/#comment-249 vlk Mon, 09 Jun 2008 01:37:08 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=308#comment-249 <em> why just b, not tb as rb in B=gs+rb or why b in S is same as in B </em> say again? why just b, not tb as rb in B=gs+rb or why b in S is same as in B

say again?

]]>
By: hlee http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/comment-page-1/#comment-246 hlee Mon, 09 Jun 2008 00:12:02 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=308#comment-246 I'm still in doubt that why just b, not tb as rb in B=gs+rb or why b in S is same as in B. How sure b is homogeneous? Nonetheless, I'm sure that the given expressions work well and homogeneity is astrophysical correct. I’m still in doubt that why just b, not tb as rb in B=gs+rb or why b in S is same as in B. How sure b is homogeneous? Nonetheless, I’m sure that the given expressions work well and homogeneity is astrophysical correct.

]]>
By: TomLoredo http://hea-www.harvard.edu/AstroStat/slog/2008/eotw-background-subtraction/comment-page-1/#comment-234 TomLoredo Fri, 23 May 2008 03:45:32 +0000 http://hea-www.harvard.edu/AstroStat/slog/?p=308#comment-234 Vinay: <em>"...old habits die hard, and old codes die harder...</em> Ouch! I think you should copyright that! 8-) Vinay:

“…old habits die hard, and old codes die harder…

Ouch! I think you should copyright that! 8-)

]]>