{ "id": 70725, "title": "Flux to magnitude", "url": "/competitions/PLAsTiCC-2018/discussion/70725", "author_name": "", "post_date": "2018-11-06T20:32:15.679382200Z", "votes": 10, "comment_count": 24, "views": 0, "content": "

Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?

\n\n

The reason I'm asking is, some variable objects can be classified on the basis of their absolute magnitude in the maximum (which is a hypothetical magnitude of an object in a normalized distance). It would obviously work only for objects with the known distance (i.e. red shift), but still it might be helpful.

", "messages": [ { "id": "416530", "postDate": "11/06/2018 20:32:15", "content": "

Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?

\n\n

The reason I'm asking is, some variable objects can be classified on the basis of their absolute magnitude in the maximum (which is a hypothetical magnitude of an object in a normalized distance). It would obviously work only for objects with the known distance (i.e. red shift), but still it might be helpful.

", "rawMarkdown": "Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?\n\nThe reason I'm asking is, some variable objects can be classified on the basis of their absolute magnitude in the maximum (which is a hypothetical magnitude of an object in a normalized distance). It would obviously work only for objects with the known distance (i.e. red shift), but still it might be helpful.", "votes": null }, { "id": "416558", "postDate": "11/06/2018 21:26:25", "content": "

Please disregard

", "rawMarkdown": "Please disregard", "votes": null }, { "id": "416714", "postDate": "11/07/2018 06:32:19", "content": "

I don't think there is any relationship between flux and magnitude given the flux is a difference between what is measured now and a reference level.

", "rawMarkdown": "I don't think there is any relationship between flux and magnitude given the flux is a difference between what is measured now and a reference level.", "votes": null }, { "id": "416740", "postDate": "11/07/2018 07:44:12", "content": "
\n

Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?

\n
\n\n

@Michal, In some cases: YES, however, it differs from Pogson magnitude by a constant.

", "rawMarkdown": ">Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?\n\n@Michal, In some cases: YES, however, it differs from Pogson magnitude by a constant.", "votes": null }, { "id": "416742", "postDate": "11/07/2018 07:49:21", "content": "
\n

it differs from Pogson magnitude by a constant.

\n
\n\n

Issue is that the constant is not constant over all sources.

", "rawMarkdown": "> it differs from Pogson magnitude by a constant.\n\nIssue is that the constant is not constant over all sources.", "votes": null }, { "id": "416743", "postDate": "11/07/2018 07:53:57", "content": "

@CPMP, talking about a constant I talk about a value constant for a given passband, not a given object.

", "rawMarkdown": "CPMP, talking about a constant I talk about a value constant for a given passband, not a given object.", "votes": null }, { "id": "416744", "postDate": "11/07/2018 07:57:58", "content": "

I agree that each passband has a constant efficiency.

\n\n

However, flux is a difference with a reference image. It means that flux is not related to magnitude unless the reference image had black background at the source location. The brightness at source location int he reference image is that constant I am speaking about. It depends on what is in the reference image, and we don't have access to that reference image unfortunately. I wonder why.

", "rawMarkdown": "I agree that each passband has a constant efficiency.\n\nHowever, flux is a difference with a reference image. It means that flux is not related to magnitude unless the reference image had black background at the source location. The brightness at source location int he reference image is that constant I am speaking about. It depends on what is in the reference image, and we don't have access to that reference image unfortunately. I wonder why.", "votes": null }, { "id": "416749", "postDate": "11/07/2018 08:09:25", "content": "

@CPMP, Imagine the cases for which the background correction used in magnitude calculations is almost equal to the \"correction\" used in LSST. These cases are the cases, for which I said YES. Talking about a constant I talk not only about different efficiencies of the detectors, but about all the effects, together with the unknown flux of the reference object of magnitude equal 0 in maximum for a given passband.

", "rawMarkdown": "CPMP, Imagine the cases for which the background correction used in magnitude calculations is almost equal to the \"correction\" used in LSST. These cases are the cases, for which I said YES. Talking about a constant I talk not only about different efficiencies of the detectors, but about all the effects, together with the unknown flux of the reference object of magnitude equal 0 in maximum for a given passband.", "votes": null }, { "id": "416763", "postDate": "11/07/2018 08:43:28", "content": "

I don't understand what you are discussing really.

\n\n

According to the document shared with us:

\n\n
\n

These light curves are the result of difference imaging: two images are taken of the same region on different nights, and the images are subtracted from each other. This differencing procedure catches both moving objects (asteroids) that are discarded for this challenge, and also objects that stay at the same position but vary in brightness. Depending on when the object first exploded or brightened, the flux may be increasing or decreasing with time.

\n
\n\n

We are only given the difference, hence I don't see how one could ever get the magnitude from it.

", "rawMarkdown": "I don't understand what you are discussing really. \n\nAccording to the document shared with us:\n\n> These light curves are the result of difference imaging: two images are taken of the same region on different nights, and the images are subtracted from each other. This differencing procedure catches both moving objects (asteroids) that are discarded for this challenge, and also objects that stay at the same position but vary in brightness. Depending on when the object first exploded or brightened, the flux may be increasing or decreasing with time.\n\nWe are only given the difference, hence I don't see how one could ever get the magnitude from it.", "votes": null }, { "id": "416767", "postDate": "11/07/2018 08:57:51", "content": "

Thanks for pointing that out, I wasn't aware of this fact. In that case this idea is a dead end.

", "rawMarkdown": "Thanks for pointing that out, I wasn't aware of this fact. In that case this idea is a dead end.", "votes": null }, { "id": "416768", "postDate": "11/07/2018 08:58:32", "content": "
\n

I don't understand what you are discussing really.

\n
\n\n

@CPMP, My simple model without this approach is worse by ~0.15, so understanding this may give you another gain of ~0.07

", "rawMarkdown": ">I don't understand what you are discussing really. \n\n@CPMP, My simple model without this approach is worse by ~0.15, so understanding this may give you another gain of ~0.07", "votes": null }, { "id": "416788", "postDate": "11/07/2018 09:27:38", "content": "

I'll try to understand what you mean, and report if anything good for me comes out of it ;)

", "rawMarkdown": "I'll try to understand what you mean, and report if anything good for me comes out of it ;)", "votes": null }, { "id": "416790", "postDate": "11/07/2018 09:28:55", "content": "

Well don't rely solely on my opinion, others like Grzegorz say they have found a way to use magnitude.

", "rawMarkdown": "Well don't rely solely on my opinion, others like Grzegorz say they have found a way to use magnitude.", "votes": null }, { "id": "416834", "postDate": "11/07/2018 10:40:44", "content": "

@Grzegorz Sionkowski I sense there is a hidden knowledge in your words. The main problem (for me) stays, how to identify such cases among all others.

", "rawMarkdown": "Grzegorz Sionkowski I sense there is a hidden knowledge in your words. The main problem (for me) stays, how to identify such cases among all others.", "votes": null }, { "id": "417195", "postDate": "11/07/2018 23:22:10", "content": "

Grzegorz,

\n\n

Not sure this is what you discussed, but I got some improvement by using squared distance in my features. Thanks for the discussion, it made me think about new things.

\n\n

I actually use redshift (hostgal_photoz) as it is almost proportional to distance AFAIK.

\n\n

I now have a single lgb model with:

\n\n

CV 0.497 LB 0.927 Gap 0.43

", "rawMarkdown": "Grzegorz,\n\nNot sure this is what you discussed, but I got some improvement by using squared distance in my features. Thanks for the discussion, it made me think about new things.\n\nI actually use redshift (hostgal_photoz) as it is almost proportional to distance AFAIK.\n\nI now have a single lgb model with:\n\nCV 0.497 LB 0.927 Gap 0.43", "votes": null }, { "id": "417468", "postDate": "11/08/2018 10:26:35", "content": "

Hey @CPMP,

\n\n

If you want to build new features by using a notion of distance, don't you think that distmod is more appropriate that hostgal_photoz?

", "rawMarkdown": "Hey @CPMP,\n\nIf you want to build new features by using a notion of distance, don't you think that `distmod` is more appropriate that `hostgal_photoz`?", "votes": null }, { "id": "417547", "postDate": "11/08/2018 12:43:39", "content": "

There is a difference between the two. distmod is a log transform of the distance. redshift is roughly proportional to distance.

", "rawMarkdown": "There is a difference between the two. distmod is a log transform of the distance. redshift is roughly proportional to distance.", "votes": null }, { "id": "419249", "postDate": "11/11/2018 15:33:31", "content": "

@CPMP what means \"squared distance\"? Did you get it from hostgal_photoz?

", "rawMarkdown": "CPMP what means \"squared distance\"? Did you get it from hostgal_photoz?", "votes": null }, { "id": "419284", "postDate": "11/11/2018 16:24:15", "content": "

I used squared hostgal_photoz.

", "rawMarkdown": "I used squared hostgal_photoz.", "votes": null }, { "id": "419417", "postDate": "11/11/2018 23:40:32", "content": "

If one basically use tree based algorithms can squaring a feature give additional gain instead of using the original feature? @CPMP

", "rawMarkdown": "If one basically use tree based algorithms can squaring a feature give additional gain instead of using the original feature? @CPMP", "votes": null }, { "id": "419425", "postDate": "11/12/2018 00:06:32", "content": "

Squaring a feature does not change much for tree based models.

\n\n

I wrote ' I got some improvement by using squared distance in my features'. I did not write that I used squared distance as a feature.

\n\n

I can't be more explicit without giving the feature(s) I am using, sorry for that.

", "rawMarkdown": "Squaring a feature does not change much for tree based models. \n\nI wrote ' I got some improvement by using squared distance in my features'. I did not write that I used squared distance as a feature.\n\nI can't be more explicit without giving the feature(s) I am using, sorry for that.", "votes": null }, { "id": "419500", "postDate": "11/12/2018 04:29:01", "content": "

No it's ok, thanks for clearing that :)

", "rawMarkdown": "No it's ok, thanks for clearing that :)", "votes": null }, { "id": "419754", "postDate": "11/12/2018 14:14:10", "content": "

is stellar magnitude same as apparent magnitude?\n@sionek, what do you mean by background correction used in magnitude calculations?

", "rawMarkdown": "is stellar magnitude same as apparent magnitude?\n@sionek, what do you mean by background correction used in magnitude calculations?", "votes": null }, { "id": "420357", "postDate": "11/13/2018 13:44:31", "content": "

@dylonLL, For a chemist using UV-Vis and FTIR spectrophotometers, background correction seems to be obvious in any kind of photometry. So, I guess, astronomers also use it, but in more sophisticated way.

\n\n

Imagine two digital pictures of a star in the form of 3x3 pixel matrix:

\n\n

1 1 1

\n\n

1 2 1

\n\n

1 1 1

\n\n

#

\n\n

2 2 2

\n\n

2 6 2

\n\n

2 2 2

\n\n

A star is in the middle (point X,Y= 2,2), and the surrounding area may be treated as a background. In a simple approach, the background is equal 1 and 2 at the first and the second picture, respectively. Removing (subtracting) the background is very important, because the brightness of the star at the picture two is not 3 times higher than at the picture one (6/2) but 4 times higher: (6-2)/(2-1). Using background correction in such meaning, you can compare not only changes of brightness of one star, but also the brightness of different stars, what is the base of stellar magnitude.

", "rawMarkdown": "dylonLL, For a chemist using UV-Vis and FTIR spectrophotometers, background correction seems to be obvious in any kind of photometry. So, I guess, astronomers also use it, but in more sophisticated way.\n\nImagine two digital pictures of a star in the form of 3x3 pixel matrix:\n\n1 1 1\n\n1 2 1\n\n1 1 1\n\n\n###\n\n\n2 2 2\n\n2 6 2\n\n2 2 2 \n\nA star is in the middle (point X,Y= 2,2), and the surrounding area may be treated as a background. In a simple approach, the background is equal 1 and 2 at the first and the second picture, respectively. Removing (subtracting) the background is very important, because the brightness of the star at the picture two is not 3 times higher than at the picture one (6/2) but 4 times higher: (6-2)/(2-1). Using background correction in such meaning, you can compare not only changes of brightness of one star, but also the brightness of different stars, what is the base of stellar magnitude.", "votes": null }, { "id": "420372", "postDate": "11/13/2018 14:10:36", "content": "

@sionek, thank you for your explanation. I am trying to extract features based on flux and meta data and your insights are very helpful

", "rawMarkdown": "sionek, thank you for your explanation. I am trying to extract features based on flux and meta data and your insights are very helpful", "votes": null } ], "comments": [ { "id": 416558, "author_name": "authman", "author_url": "", "post_date": "11/06/2018 21:26:25", "content": "

Please disregard

", "votes": null, "replies": [] }, { "id": 416714, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 06:32:19", "content": "

I don't think there is any relationship between flux and magnitude given the flux is a difference between what is measured now and a reference level.

", "votes": null, "replies": [ { "id": 416767, "author_name": "rejpalcz", "author_url": "", "post_date": "11/07/2018 08:57:51", "content": "

Thanks for pointing that out, I wasn't aware of this fact. In that case this idea is a dead end.

", "votes": null, "replies": [] }, { "id": 416790, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 09:28:55", "content": "

Well don't rely solely on my opinion, others like Grzegorz say they have found a way to use magnitude.

", "votes": null, "replies": [] } ] }, { "id": 416740, "author_name": "sionek", "author_url": "", "post_date": "11/07/2018 07:44:12", "content": "
\n

Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?

\n
\n\n

@Michal, In some cases: YES, however, it differs from Pogson magnitude by a constant.

", "votes": null, "replies": [ { "id": 416742, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 07:49:21", "content": "
\n

it differs from Pogson magnitude by a constant.

\n
\n\n

Issue is that the constant is not constant over all sources.

", "votes": null, "replies": [] }, { "id": 416743, "author_name": "sionek", "author_url": "", "post_date": "11/07/2018 07:53:57", "content": "

@CPMP, talking about a constant I talk about a value constant for a given passband, not a given object.

", "votes": null, "replies": [] }, { "id": 416744, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 07:57:58", "content": "

I agree that each passband has a constant efficiency.

\n\n

However, flux is a difference with a reference image. It means that flux is not related to magnitude unless the reference image had black background at the source location. The brightness at source location int he reference image is that constant I am speaking about. It depends on what is in the reference image, and we don't have access to that reference image unfortunately. I wonder why.

", "votes": null, "replies": [] }, { "id": 416749, "author_name": "sionek", "author_url": "", "post_date": "11/07/2018 08:09:25", "content": "

@CPMP, Imagine the cases for which the background correction used in magnitude calculations is almost equal to the \"correction\" used in LSST. These cases are the cases, for which I said YES. Talking about a constant I talk not only about different efficiencies of the detectors, but about all the effects, together with the unknown flux of the reference object of magnitude equal 0 in maximum for a given passband.

", "votes": null, "replies": [] }, { "id": 416763, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 08:43:28", "content": "

I don't understand what you are discussing really.

\n\n

According to the document shared with us:

\n\n
\n

These light curves are the result of difference imaging: two images are taken of the same region on different nights, and the images are subtracted from each other. This differencing procedure catches both moving objects (asteroids) that are discarded for this challenge, and also objects that stay at the same position but vary in brightness. Depending on when the object first exploded or brightened, the flux may be increasing or decreasing with time.

\n
\n\n

We are only given the difference, hence I don't see how one could ever get the magnitude from it.

", "votes": null, "replies": [] }, { "id": 416768, "author_name": "sionek", "author_url": "", "post_date": "11/07/2018 08:58:32", "content": "
\n

I don't understand what you are discussing really.

\n
\n\n

@CPMP, My simple model without this approach is worse by ~0.15, so understanding this may give you another gain of ~0.07

", "votes": null, "replies": [] }, { "id": 416788, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 09:27:38", "content": "

I'll try to understand what you mean, and report if anything good for me comes out of it ;)

", "votes": null, "replies": [] }, { "id": 416834, "author_name": "rejpalcz", "author_url": "", "post_date": "11/07/2018 10:40:44", "content": "

@Grzegorz Sionkowski I sense there is a hidden knowledge in your words. The main problem (for me) stays, how to identify such cases among all others.

", "votes": null, "replies": [] }, { "id": 417195, "author_name": "cpmpml", "author_url": "", "post_date": "11/07/2018 23:22:10", "content": "

Grzegorz,

\n\n

Not sure this is what you discussed, but I got some improvement by using squared distance in my features. Thanks for the discussion, it made me think about new things.

\n\n

I actually use redshift (hostgal_photoz) as it is almost proportional to distance AFAIK.

\n\n

I now have a single lgb model with:

\n\n

CV 0.497 LB 0.927 Gap 0.43

", "votes": null, "replies": [] }, { "id": 417468, "author_name": "maxhalford", "author_url": "", "post_date": "11/08/2018 10:26:35", "content": "

Hey @CPMP,

\n\n

If you want to build new features by using a notion of distance, don't you think that distmod is more appropriate that hostgal_photoz?

", "votes": null, "replies": [] }, { "id": 417547, "author_name": "cpmpml", "author_url": "", "post_date": "11/08/2018 12:43:39", "content": "

There is a difference between the two. distmod is a log transform of the distance. redshift is roughly proportional to distance.

", "votes": null, "replies": [] }, { "id": 419249, "author_name": "suvalex", "author_url": "", "post_date": "11/11/2018 15:33:31", "content": "

@CPMP what means \"squared distance\"? Did you get it from hostgal_photoz?

", "votes": null, "replies": [] }, { "id": 419284, "author_name": "cpmpml", "author_url": "", "post_date": "11/11/2018 16:24:15", "content": "

I used squared hostgal_photoz.

", "votes": null, "replies": [] }, { "id": 419417, "author_name": "rahulsrk", "author_url": "", "post_date": "11/11/2018 23:40:32", "content": "

If one basically use tree based algorithms can squaring a feature give additional gain instead of using the original feature? @CPMP

", "votes": null, "replies": [] }, { "id": 419425, "author_name": "cpmpml", "author_url": "", "post_date": "11/12/2018 00:06:32", "content": "

Squaring a feature does not change much for tree based models.

\n\n

I wrote ' I got some improvement by using squared distance in my features'. I did not write that I used squared distance as a feature.

\n\n

I can't be more explicit without giving the feature(s) I am using, sorry for that.

", "votes": null, "replies": [] }, { "id": 419500, "author_name": "rahulsrk", "author_url": "", "post_date": "11/12/2018 04:29:01", "content": "

No it's ok, thanks for clearing that :)

", "votes": null, "replies": [] }, { "id": 419754, "author_name": "niclasdoce", "author_url": "", "post_date": "11/12/2018 14:14:10", "content": "

is stellar magnitude same as apparent magnitude?\n@sionek, what do you mean by background correction used in magnitude calculations?

", "votes": null, "replies": [] }, { "id": 420357, "author_name": "sionek", "author_url": "", "post_date": "11/13/2018 13:44:31", "content": "

@dylonLL, For a chemist using UV-Vis and FTIR spectrophotometers, background correction seems to be obvious in any kind of photometry. So, I guess, astronomers also use it, but in more sophisticated way.

\n\n

Imagine two digital pictures of a star in the form of 3x3 pixel matrix:

\n\n

1 1 1

\n\n

1 2 1

\n\n

1 1 1

\n\n

#

\n\n

2 2 2

\n\n

2 6 2

\n\n

2 2 2

\n\n

A star is in the middle (point X,Y= 2,2), and the surrounding area may be treated as a background. In a simple approach, the background is equal 1 and 2 at the first and the second picture, respectively. Removing (subtracting) the background is very important, because the brightness of the star at the picture two is not 3 times higher than at the picture one (6/2) but 4 times higher: (6-2)/(2-1). Using background correction in such meaning, you can compare not only changes of brightness of one star, but also the brightness of different stars, what is the base of stellar magnitude.

", "votes": null, "replies": [] }, { "id": 420372, "author_name": "niclasdoce", "author_url": "", "post_date": "11/13/2018 14:10:36", "content": "

@sionek, thank you for your explanation. I am trying to extract features based on flux and meta data and your insights are very helpful

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "416530": "Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?\n\nThe reason I'm asking is, some variable objects can be classified on the basis of their absolute magnitude in the maximum (which is a hypothetical magnitude of an object in a normalized distance). It would obviously work only for objects with the known distance (i.e. red shift), but still it might be helpful.", "416558": "Please disregard", "416714": "I don't think there is any relationship between flux and magnitude given the flux is a difference between what is measured now and a reference level.", "416740": ">Do you think there exists (and we can discover) a relationship between a flux in this competition and stellar (Pogson) magnitude?\n\n@Michal, In some cases: YES, however, it differs from Pogson magnitude by a constant.", "416742": "> it differs from Pogson magnitude by a constant.\n\nIssue is that the constant is not constant over all sources.", "416743": "CPMP, talking about a constant I talk about a value constant for a given passband, not a given object.", "416744": "I agree that each passband has a constant efficiency.\n\nHowever, flux is a difference with a reference image. It means that flux is not related to magnitude unless the reference image had black background at the source location. The brightness at source location int he reference image is that constant I am speaking about. It depends on what is in the reference image, and we don't have access to that reference image unfortunately. I wonder why.", "416749": "CPMP, Imagine the cases for which the background correction used in magnitude calculations is almost equal to the \"correction\" used in LSST. These cases are the cases, for which I said YES. Talking about a constant I talk not only about different efficiencies of the detectors, but about all the effects, together with the unknown flux of the reference object of magnitude equal 0 in maximum for a given passband.", "416763": "I don't understand what you are discussing really. \n\nAccording to the document shared with us:\n\n> These light curves are the result of difference imaging: two images are taken of the same region on different nights, and the images are subtracted from each other. This differencing procedure catches both moving objects (asteroids) that are discarded for this challenge, and also objects that stay at the same position but vary in brightness. Depending on when the object first exploded or brightened, the flux may be increasing or decreasing with time.\n\nWe are only given the difference, hence I don't see how one could ever get the magnitude from it.", "416767": "Thanks for pointing that out, I wasn't aware of this fact. In that case this idea is a dead end.", "416768": ">I don't understand what you are discussing really. \n\n@CPMP, My simple model without this approach is worse by ~0.15, so understanding this may give you another gain of ~0.07", "416788": "I'll try to understand what you mean, and report if anything good for me comes out of it ;)", "416790": "Well don't rely solely on my opinion, others like Grzegorz say they have found a way to use magnitude.", "416834": "Grzegorz Sionkowski I sense there is a hidden knowledge in your words. The main problem (for me) stays, how to identify such cases among all others.", "417195": "Grzegorz,\n\nNot sure this is what you discussed, but I got some improvement by using squared distance in my features. Thanks for the discussion, it made me think about new things.\n\nI actually use redshift (hostgal_photoz) as it is almost proportional to distance AFAIK.\n\nI now have a single lgb model with:\n\nCV 0.497 LB 0.927 Gap 0.43", "417468": "Hey @CPMP,\n\nIf you want to build new features by using a notion of distance, don't you think that `distmod` is more appropriate that `hostgal_photoz`?", "417547": "There is a difference between the two. distmod is a log transform of the distance. redshift is roughly proportional to distance.", "419249": "CPMP what means \"squared distance\"? Did you get it from hostgal_photoz?", "419284": "I used squared hostgal_photoz.", "419417": "If one basically use tree based algorithms can squaring a feature give additional gain instead of using the original feature? @CPMP", "419425": "Squaring a feature does not change much for tree based models. \n\nI wrote ' I got some improvement by using squared distance in my features'. I did not write that I used squared distance as a feature.\n\nI can't be more explicit without giving the feature(s) I am using, sorry for that.", "419500": "No it's ok, thanks for clearing that :)", "419754": "is stellar magnitude same as apparent magnitude?\n@sionek, what do you mean by background correction used in magnitude calculations?", "420357": "dylonLL, For a chemist using UV-Vis and FTIR spectrophotometers, background correction seems to be obvious in any kind of photometry. So, I guess, astronomers also use it, but in more sophisticated way.\n\nImagine two digital pictures of a star in the form of 3x3 pixel matrix:\n\n1 1 1\n\n1 2 1\n\n1 1 1\n\n\n###\n\n\n2 2 2\n\n2 6 2\n\n2 2 2 \n\nA star is in the middle (point X,Y= 2,2), and the surrounding area may be treated as a background. In a simple approach, the background is equal 1 and 2 at the first and the second picture, respectively. Removing (subtracting) the background is very important, because the brightness of the star at the picture two is not 3 times higher than at the picture one (6/2) but 4 times higher: (6-2)/(2-1). Using background correction in such meaning, you can compare not only changes of brightness of one star, but also the brightness of different stars, what is the base of stellar magnitude.", "420372": "sionek, thank you for your explanation. I am trying to extract features based on flux and meta data and your insights are very helpful" }, "source": "meta" }