[vsnet-alert 11046] re Malmquist Bias reference

[varposts at Safe-mail.net]

Well, Wolfgang Renz was kind enough to let me know that the reference was to a very obscure paper in 1920 and it was about A stars apparently, not galaxies after all, with the bias being later co-opted by galaxy researchers for their usage.  It seems there was some sort of statistics' school known as the "Lund Group" in Swedish astronomy in them days and a few things came out of that.

Malmquist Bias is about detection and nondetection within a sample giving a bias selection effect, basically.  My analogy was rubbishy so it seemed to reinforce misconceptions that arise because of late it's been co-opted into effects in surveys deemed due to instrumentation.  It isn't the instrumentation at fault, it's an inherent principle is this.

The magnitude of an object is measured by receipt of a bunch of photons of varying intensity (usually sampled over some short range of the "continuum", and of course usually having traversed our atmosphere).  Near, and at, limit detection the low intensity stuff just doesn't get detected (or has a different probability of being detected), whereas the high does (or has a higher probability of being detected), you don't get a normal distribution of detections, you get the 'high' end, for the simple case where an object is of a magnitude that straddles this detection limit, and as only part is sampled the measure ends up artifically bright.  Images are an extended range of flux measurements on a substrate / detector array that are summed, remember.  Surveys at their limits of detection can return falsely bright magnitudes for objects, with the extreme cases being outlined in Brian Skiff's post, ie where a nondetection can actually be given as a real detection if just data from the mag column is used.  So a "brightness" selection effect can be forced into consideration with detectors, but it's not about detectors.

When SDSS tells you they've got galaxies measured down to mag 24, you know some of those are really mag 26 or 25 or whatever, so the mag 24 stuff isn't a clean sample, and Malmquist Bias has to be considered when you select the limit for a sample to study.  Nondetected stuff appearing real depends on how they go about things in a particular survey.

So you simply don't use survey data down to the limit.  You've gotta know a survey's limitations...  And with 2MASS you've got to remember that Ks' limitation is around a mag higher than J and H.  You'll see 16th to 17th mag J values quoted around though, without quoted errors, and probably without flag checking.

You can force 'Malmquist Bias'-esque aspects into faint limit, and I have seen it expressed solely in terms of being some instrumental flux thing, and as the original reference is not readily accessible folk probably thinks that's where it hails from.  There's a lot of astronomy and general principles been around a long time, way before CCDs, that may apply to CCDs but weren't found because of CCDs, or any other detector for that matter.

Basically, I should've just said "I didn't mention 2MASS colours as the 2MASS K band for this object is pretty much rubbish so pretty meaningless" and had done with it, there was a chance folk'd've dismissed as John just being stroppy again and not got the point, but at least it'd've been accurate.  2MASS colours on the whole won't be much use for Catalina CV candidates coz of the mags they usually have at quiescence, and surveys are artificially bright biased at their faint ends.  Just coz something's in a catalogue don't mean it's right, it depends, and also catalogues nowadays often kick around other extra info to help you make decisions, they don't try to con folk in anyway, it's folk that cons themselves.

Cheers

John