Asymmetrical Subset 2 (QS)

by ChrisMolloy
I've started looking at the QS Asymmetrical galaxies in subset 2 in excel. I have some questions on the classifications relating to the core data.
Tentatively I have approximately 270 asymmetrical galaxies with about 127 in the Neither category. I'm classifying initially all Asymmetrical with a vote of 10 or greater.
Cllassifications with a split 10/10 I've included as Asymmetrical. Is it correct to include a 10/10 as asymmetrical? I've also wondered about the 11/9's 10/9's and 10/8's. And also vice versa such as the 8/10's, /9/11's. etc. There are some 8/8's and 9/9's too. So, some numbers don't add up to 20 as I'm sure you know. Have been doing quite a lot of visual inspections and found that helpful. What do we do here?
Secondly, using Simbad, how do I find if any of the asymmetrical are in clusters or not? Have looked but are a bit lost. As an example getting a list of 20 galaxies. What does this mean? How do I work with this? Is there another way? Usually in GZ I go to Simbad and can see a galaxy listed as being part of a cluster/group but I'm a bit flummoxed at what I'm seeing and more importantly how to interpret.
Thirdly , what to do if a classification appears incorrect. Such as classified as Smooth when NED says IRS? However, could be "visually" Smooth so then the question is how do we properly classify it? Smooth/FOD?
Posted

by mlpeck in response to ChrisMolloy's comment.
Why not try different thresholds and see if it makes a difference? See the merger thread that JeanTate started recently.
Posted

by JeanTate in response to ChrisMolloy's comment.
I agree with mlpeck's suggestion, try different thresholds, and see what patterns you get.
Cllassifications with a split 10/10 I've included as Asymmetrical. Is it correct to include a 10/10 as asymmetrical? I've also wondered about the 11/9's 10/9's and 10/8's. And also vice versa such as the 8/10's, /9/11's. etc. There are some 8/8's and 9/9's too. So, some numbers don't add up to 20 as I'm sure you know. Have been doing quite a lot of visual inspections and found that helpful. What do we do here?
For subsample_2 QS objects, there's perhaps just one object whose soa vote fraction is large enough to warrant serious consideration. For the rest, unless someone has a good reason to do otherwise, I'd simply take the vote fractions of the nonsoa classifications: 10/10 > 0.5, 11/9 > 0.55, 8/8 > 0.5, 10/8 > 0.56 (to two significant figures), etc.
Secondly, using Simbad, how do I find if any of the asymmetrical are in clusters or not?
I'll have to pass on that ... I've hardly ever used Simbad ðŸ˜®
Thirdly , what to do if a classification appears incorrect.
Make a note/keep a list, and send it to me (AGS IDs, plus anything you want to say about them), or post a list of them, in this thread for example! However, if the analysis you're planning on doing is 'asymmetric fraction by log_mass by QS vs QC'  or something like this  then it won't matter whether the Smooth/Fod classifications are correct or not, will it?
Posted

by ChrisMolloy in response to JeanTate's comment.
I'll adjust the thresholds as mlpeck and yourself have suggested.
Thirdly , what to do if a classification appears incorrect. Make a note/keep a list, and send it to me (AGS IDs, plus anything you want to say about them), or post a list of them, in this thread for example! However, if the analysis you're planning on doing is 'asymmetric fraction by log_mass by QS vs QC'  or something like this  then it won't matter whether the Smooth/Fod classifications are correct or not, will it?
I'll keep a list and post it here. I may look at some differences between Smooth/FoD.
Posted

by ChrisMolloy
Hereâ€™s a brief look at the QS Asymmetrical Merging and Neither categories by log mass and redshift.
The threshold I examined is .5. The merging category is defined as Merger, Tidal Debris and Both. Total asymmetrical galaxies are 281 which include 143 Mergers and 138 Neither.
Where there are no data points in some of the graphs there are no galaxies at those log masses. Six log mass bins were used which are 010, 1010.3, 10.310.6, 10.610.9, 10.911.2, > 11.2. Both categories are examined in 4 redshift bins, z=0204, z=0406, z=0608, z=081, as well as an overview of the log mass below.
The Merging category rises from a fraction of 6 at log mass 9.84 to a fraction of 8 at log mass 10.45. It then rises steeply to a fraction of 28 at log mass 10.75, fraction of 36 at log mass 11.01, and a fraction of 50 at log mass 11.41. The Neither category is more of a gentle wave, with not much difference in fraction between the log mass bins; a fraction of 13 at log mass 9.66, ending on a fraction of 14 at log mass 11.01.
Redshift z0204 doesnâ€™t have as much data but the trend is slightly similar to the above overview. A wave type function for Neither, and a rise at log mass 10.4 for Merging.
Redshift z=0406 shows a steep ascent for the Merging galaxies; from a fraction of 2 at log mass 9.99 to a fraction of 50 at log mass 10.93. There is more of a dip for the Neither galaxies, beginning at a fraction of 17 at log mass 9.65, dropping to a fraction of 1at log mass 10.10, and rising and levelling off at a fraction of 17 at log mass 11.00.
Redshift z=0608 the Neither category again has a wave type function. It begins at a fraction of 14 at log mass 9.66, dropping to a fraction of 8 at log mass 10.46, rising to a fraction of 14 at log mass 10.74 and dipping to a fraction of 10 at log mass 10.97. The Merging category begins at a fraction of 10 at log mass 9.84, dips to a fraction of 5 at log mass 10.49 and then rises steeply to a fraction of 40 at log mass 11.04.
Redshift z=081 the Merging category is more step like, but gradually increasing from a fraction of 9 at log mass 10.46 to a fraction of 50 at log mass 11.41. The Neither category rises from a fraction of 10 at log mass 9.78, peaking at a fraction of 19 at log mass 10.75 and then dipping to a fraction of 17 at log mass 11.02.
Summary
In all cases the log mass fraction value for the Neither category does not exceed 20. For the merging category the fraction value generally exceeds a value of 40. At log mass 10.4 â€“ 10.5 the fraction for merging increases dramatically with log mass. While for the Neither category this does occur but is less pronounced and levels off or decreases.Posted

by JeanTate in response to ChrisMolloy's comment.
Cool! ðŸ˜ƒ
Some questions on clarification, if I may, to be sure that I've understood your results correctly; in a later post I'll ask about the content.
a brief look at the QS Asymmetrical Merging and Neither categories by log mass and redshift
Your 'universe' (or 'population'  is that the correct technical word, anyone?) is "the 1149", QS galaxies with redshifts between 0.02 and 0.10 AND with zband estimated absolute magnitudes of 21.0 or brighter, right?
The threshold I examined is .5
There are two of these. One is the 'asymmetry fraction'; for a particular galaxy, if 't10_a01_count'/('t10_a01_count' + 't10_a00_count') is greater than or equal to 0.5, it makes the cut. The other is the 'merger fraction'; dropping the 't09_' and '_count', the number is ('a01'+'a02'+'a03')/('a00'+'a01'+'a02'+'a03'). Right?
Six log mass bins were used which are 010, 1010.3, 10.310.6, 10.610.9, 10.911.2, > 11.2
Not that any galaxy has a log_mass exactly equal to 10, 10.3, 10.6, or 10.9  so my question is moot in a practical sense  into which bins would such galaxies fall? For example, if a galaxy's log_mass were 10.0000... , would it belong to the first bin or the second?
The dataset you used is the one in which all the 'missing mass' (i.e. value of 1) values were replaced by the values mlpeck reported, right?
in 4 redshift bins, z=0204, z=0406, z=0608, z=081
This is shorthand for 0.02, 0.04, ... 0.10, right? Same question re which bin an object with a 'boundary value' falls in.
The Merging category rises from a fraction of 6 at log mass 9.84 to a fraction of 8 at log mass 10.45
The "6" means 6%, right? And "9.84" is the median log_mass of the first log_mass bin, right? (Similarly for all others).
Your plots all have xaxes labeled as "Log Mass Median"; why median (and not, say, mean)?
What is the yaxis ("Fraction")? For example, for the first green ("Merger") point  (9.84, 6)  I think it's '(number of QS objects in the first log_mass bin with a merger fraction >= 0.5)/('total number of QS objects in the first log_mass bin').
Posted

by ChrisMolloy in response to JeanTate's comment.
Thanks for the reply.
Your 'universe' (or 'population'  is that the correct technical word, anyone?) is "the 1149", QS galaxies with redshifts between 0.02 and 0.10 AND with zband estimated absolute magnitudes of 21.0 or brighter, right?
Everything above is correct except the following. Mine are zband estimated absolute magnitudes of 20.0 or brighter. I must have missed a post somewhere to filter for 21.0 or brighter. In your post here you reference absolute magnitudes of 20.0 or brighter. Why the difference?
There are two of these. One is the 'asymmetry fraction'; for a particular galaxy, if 't10_a01_count'/('t10_a01_count' + 't10_a00_count') is greater than or equal to 0.5, it makes the cut. The other is the 'merger fraction'; dropping the 't09_' and '_count', the number is ('a01'+'a02'+'a03')/('a00'+'a01'+'a02'+'a03'). Right?
Yes. Anything .5 or greater makes the cut. I did this slightly differently. For asymmetry (total) I filtered T010 a01 as greater than equal to 10. This gave me 270 galaxies. I then filtered for .5 thresholds less than 10 using the equal ratio e.g.9/9, 8/8. for T010 a00 and T010 a01. This gave me an additional 11 galaxies with a grand total of 281 galaxies. However having just gone back through and checked the data I'm possibly out by 5 galaxies. I excluded the 9/8 and 8/7 ratios.
So for Mergers, t09 a03 was filtered as less than or equal to 9 and T010 a01 as greater than or egual to 10. This gave me 143 galaxies. For Neither I fitered T010 a01 as greater than equal to 10 and T09 a03 as greater than or equal to 10. This gave me 127 galaxies. The difference was made up by the additional 11 galaxies being added to the Neither category but that may be overestimated by 2 galaxies.
Overall, I need to tweak these numbers a bit more. Also zband estimated absolute magnitudes of 21.0 or brighter will damatically change the above plots.
Not that any galaxy has a log_mass exactly equal to 10, 10.3, 10.6, or 10.9  so my question is moot in a practical sense  into which bins would such galaxies fall? For example, if a galaxy's log_mass were 10.0000... , would it belong to the first bin or the second?
10.000 in the second bin. So if you had log mass 10.60000 it goes into the 10.610.9 bin.
The dataset you used is the one in which all the 'missing mass' (i.e. value of 1) values were replaced by the values mlpeck reported, right?
Correct.
in 4 redshift bins, z=0204, z=0406, z=0608, z=081 This is shorthand for 0.02, 0.04, ... 0.10, right? Same question re which bin an object with a 'boundary value' falls in.
This is correct for the shorthand. The boundary value is the same as applied to log mass above but applicable to redshift.
The Merging category rises from a fraction of 6 at log mass 9.84 to a fraction of 8 at log mass 10.45. The "6" means 6%, right? And "9.84" is the median log_mass of the first log_mass bin, right? (Similarly for all others).
The 6 does mean 6%. Yes; 9.84 is for the first bin. And again similarly for the other bins. And just to be clear the median is for the Neither and Merging galaxies in each bin.
Your plots all have xaxes labeled as "Log Mass Median"; why median (and not, say, mean)?
The median is the number in the middle of a set of numbers; that is, half the numbers have values that are greater than the median, and half have values that are less (copied that from excel help). The mean is the arithmetic average of a set of values, or distribution, (wiki). Why would you plot the meaan in this case? It could skew the bins.
What is the yaxis ("Fraction")? For example, for the first green ("Merger") point  (9.84, 6)  I think it's '(number of QS objects in the first log_mass bin with a merger fraction >= 0.5)/('total number of QS objects in the first log_mass bin').
It is the number of QS objects in the first log_mass bin with a merger fraction >= 0.5)/('total number of QS objects in the first log_mass bin').
Posted

by JeanTate in response to ChrisMolloy's comment.
Thanks! ðŸ˜ƒ
21.0 is a typo! ðŸ˜¦ It should be 20.0 (sorry).
(just wanted to note that quickly; more later)
Posted

by JeanTate in response to ChrisMolloy's comment.
Thanks again.
Yes. Anything .5 or greater makes the cut. I did this slightly differently. For asymmetry (total) I filtered T010 a01 as greater than equal to 10. This gave me 270 galaxies. I then filtered for .5 thresholds less than 10 using the equal ratio e.g.9/9, 8/8. for T010 a00 and T010 a01. This gave me an additional 11 galaxies with a grand total of 281 galaxies. However having just gone back through and checked the data I'm possibly out by 5 galaxies. I excluded the 9/8 and 8/7 ratios.
I did a quick, independent, check, and came up with 288; I'll go through this more carefully later.
Overall, I need to tweak these numbers a bit more.
If it would help, I'll do an independent check of the "Mergers" too.
The median is the number in the middle of a set of numbers; that is, half the numbers have values that are greater than the median, and half have values that are less (copied that from excel help). The mean is the arithmetic average of a set of values, or distribution, (wiki). Why would you plot the meaan in this case? It could skew the bins.
Good point. I wonder what measure is commonly used by astronomers, in this kind of study?
I guess its real importance is this: whatever measure we use, we must be consistent, in all analyses! ðŸ˜ƒ
Also, time to think about what  common, consistent  error bars to use ....
Posted

by ChrisMolloy in response to JeanTate's comment.
Thanks. I rechecked the numbers. I get 288 asymmetrical, consisting of 147 Mergers and 141 Neither. Can you cross check these? Once I here back from you I'll modify the above.
Good point. I wonder what measure is commonly used by astronomers, in this kind of study?
Can't answer that one, although I can see the validity of both the Mean and Median, depending on the circumstances.
Also, time to think about what  common, consistent  error bars to use ....
So, what do we do here?
Posted

by JeanTate in response to ChrisMolloy's comment.
Thanks. I rechecked the numbers. I get 288 asymmetrical, consisting of 147 Mergers and 141 Neither. Can you cross check these?
Hmm, I got 153 Mergers and 135 Neither. While it may be just a coincidence, I found there are six objects which have total t09_a03 counts ("Neither" to the merger question) of 10, AND which are asymmetric (i.e. count fraction >= 0.5), AND for which the sum of the other 'merger' counts (i.e. t09_a00 Merging, t09_a01 Tidal debris, and t09_a02 Both) is also 10.
Good point. I wonder what measure is commonly used by astronomers, in this kind of study?
Can't answer that one, although I can see the validity of both the Mean and Median, depending on the circumstances.
I started a thread on this, in the Questions for the Scientists! board: When binning, what's the best measure to use (mean, median, ...)? Let's see what responses we get.
Also, time to think about what  common, consistent  error bars to use ....
So, what do we do here?
Some time ago, I started a thread related to this, also in the Questions for the Scientists! board: How to derive physically real error bars, for fractions near 0 and 1? There are some interesting posts from mlpeck and various scientists! ðŸ˜ƒ I plan to explore some of those further, with the aim of finding an easytoimplementinaspreadsheet method, one that is also both standard practice among astronomers and is physically real (or very close to it). It'd be great if you could join the discussion! ðŸ˜„
Posted

by ChrisMolloy in response to JeanTate's comment.
Hmm, I got 153 Mergers and 135 Neither. While it may be just a coincidence, I found there are six objects which have total t09_a03 counts ("Neither" to the merger question) of 10, AND which are asymmetric (i.e. count fraction >= 0.5), AND for which the sum of the other 'merger' counts (i.e. t09_a00 Merging, t09_a01 Tidal debris, and t09_a02 Both) is also 10.
That is a really good example of something that's been bugging me. Question of thresholds. So, all >=0.5 will depend on what your bias is. If you filter all galaxies as symmetrical for T10 a00 on a >=0.5 you will get an overlap if you ask the same question for T10 a01 as to whether they are asymmetrical. Hence, all galaxies with an equual vote of 10/10 can go into either category.
So with regard to your example above I filtered T10 a01 and T09a03 as =>10 and those six galaxies will come up as neither asymmetrical. Then I filtered the asymmetrical down for the 9/9's 9/8's 8/7's etc. To get the mergers I fitered T10 a01 as => 10 and T09 a03 as =LT9 etc. Those galaxies will not appear in the mergers. But if you ask your question of merging, then they could appear in the asymmetrical merging category if you use a sum count of T09 a00, a01,a02 being 10 (all merger votes equals 10) with T09 a03 (neither votes) being 10 (using the threshold of >=0.5), with the merging question taking precedence. So, those six galaxies above to me would be neither asymmetrical. However, they could also be merging asymmetrical. So, what to do for consistency? I wonder how this issue was resolved in Tools or is usually settled? Do you have any views?
Some time ago, I started a thread related to this, also in the Questions for the Scientists! board: How to derive physically real error bars, for fractions near 0 and 1? There are some interesting posts from mlpeck and various scientists! ðŸ˜ƒ I plan to explore some of those further, with the aim of finding an easytoimplementinaspreadsheet method, one that is also both standard practice among astronomers and is physically real (or very close to it). It'd be great if you could join the discussion! ðŸ˜„
Have been following this thread with interest. Don't know what to say yet. And I will follow the new thread on the Mean and the Median.
Posted

by JeanTate in response to ChrisMolloy's comment.
So, what to do for consistency? I wonder how this issue was resolved in Tools or is usually settled? Do you have any views?
I don't know about Tools, but otherwise it seems  to me  like a question about how to treat 'boundary' cases. I've always tried to be both consistent and exclusive ... when dividing a set of objects into subsets, by the values of just one parameter, make sure that the rule you use/apply is unambiguous, that no object can belong to more than one subset, and that every object belongs to a subset (of course, subsets may contain zero objects).
So, for example, in the redshift bins we discussed earlier, the first bin (subset) might be 0.02 <= z < 0.04, and the second 0.04 <= z < 0.06 (etc); an object with a redshift of exactly 0.04 would belong to the second bin, not the first*.
For the various 'X fractions' ('asymmetry fraction', 'merger fraction'): treat each question (t09, t10) as independent (obviously, but it's worth stating), and first decide on an independent rule for 'outliers'. For the objects and questions we're now working with, I think the only possible 'outliers' are those which might be 'star or artifact': choose a threshold (a 'soa fraction' of 0.5, say) and remove all objects above the threshold**.
If your rule is based on fractions, not vote totals, then first define your rule unambiguously (in terms of fractions, starting with a definition), then apply it consistently.
For a threshold which is a fraction, and 0.5, there will be ambiguity, as you stated. If you change the rule/threshold to 0.6 (say), does the ambiguity go away (I think so)? What about 0.55? ... so one way to sidestep the ambiguity that comes with a threshold of 0.5 is to set the threshold to a value just above (or below!) 0.5; you get to keep consistency and exclusivity, and it's easier to wrap your head around it ðŸ˜ƒ
Does this help?
And I will follow the new thread on the Mean and the Median.
It's been quiet there ... ðŸ˜¦
*a further clarification: what to do if the measured redshift is 0.0399Â±0.0006; do you 'round up' to incorporate the stated uncertainty? Or go with the central value regardless of the 'error bars'? It seems to me the first thing is to be clear  and recognizing the potential problem is the precursor to that  then, as long as you're completely consistent, it won't matter, will it?
**I think there's only one object that comes close
Posted

by ChrisMolloy in response to JeanTate's comment.
This does help. Took me a while to get my head around what you were saying and how to apply it. I'll stick with the 0.5 threshold and amend the results as per your top post above (will be in a couple of days). The sum count of t09 a00,a01,a02 is more accurate with regards to mergers or not. And it is compatible with how the log mass & redshift boundaries were defined.
*a further clarification: what to do if the measured redshift is 0.0399Â±0.0006; do you 'round up' to incorporate the stated uncertainty? Or go with the central value regardless of the 'error bars'? It seems to me the first thing is to be clear  and recognizing the potential problem is the precursor to that  then, as long as you're completely consistent, it won't matter, will it?
That last sentence is a good point. Also I thnk there is one possible soa in my data with a threshold of =0.5. Will check this object.
Posted

by ChrisMolloy
Hereâ€™s the update of the QS asymmetrical merging and neither categories.
Galaxies examined are QS 1149 with redshifts between 0.02 and 0.10 and with zband estimated absolute magnitudes of 20.0 or brighter. Total asymmetrical galaxies are 288, which include 155 mergers and 133 neither. The merging category is defined as merger, tidal debris and both.
The threshold examined is equals or greater than 0.5. The method used for classifying asymmetrical is if 't10_a01 (asymmetrical) count is greater than or equal to 0.5 in contrast to t10_a00 (symmetrical), it is asymmetrical. To distinguish merger and neither, if the sum count of t09_a00, a01, a02 (mergers) is greater than or equal to 0.5 in contrast to t09_a03 (neither) it is a merger. Anything less than this is neither.
Where there are no data points in some of the graphs there are no galaxies at those log masses. Six log mass bins were used which are 010, 1010.3, 10.310.6, 10.610.9, 10.911.2, > 11.2. Both categories are examined in 4 redshift bins, z=0.020.04, z=0.040.06, z=0.060.08, z=0.080.10, as well as an overview of the log mass below. Nothing has been rounded and the central value has been used.
The x axis represents fractions, the y axis is the arithmetic mean. Fractions reflect the actual fraction, not percentage.
The merging category rises from a fraction of 0.06 at log mass 9.87 to a fraction of 0.09 at log mass 10.46. It then rises steeply to a fraction of 0.28 at log mass 10.75, fraction of 0.37 at log mass 10.99 and a fraction of 0.5 at log mass 11.41. The neither category is more of a gentle wave, with not much difference in the fractions between the log mass bins, fraction of 0.13 at log mass 9.8, peaking at a fraction of 0.17 at log mass 10.73 and ending on a fraction of 0.12 at log mass 11.02.
Redshift z = 0.020.04 doesnâ€™t have much data. Main features are a wave type function again for neither, and a rise at log mass 10.4 for merging, peaking at a fraction of 1 at log mass 10.92.
Redshift z = 0.04 â€“ 0.06 shows a steep ascent for the merging galaxies, from a fraction of 0.01 at log mass 9.98 to a fraction of 0.5 at log mass 10.93. There is a dip and then a rise for the neither galaxies; beginning at a fraction of 0.16 at log mass 9.8, dipping to a fraction of 0.01at log mass 10.1, and rising and levelling off at a fraction of 0.16 at log mass 10.99.
For redshift z = 0.06  0.08 the neither category again displays a slight wave type function. It begins at a fraction of 0.15 at log mass 9.8, dipping to a fraction of 0.07 at log mass 10.41, rising to a fraction of 0.11 at log mass 10.72 and dipping to a fraction of 0.1 at log mass 10.97. The merging category rises from a fraction of 0.09 at log mass 9.83 to a fraction of 0.11 at log mass 10.18, dips to a fraction of 0.07 at log mass 10.48 and then rises steeply to a fraction of 0.4 at log mass 11.02.
Redshift z = 0.08 â€“ 0.10 the merging category is more step like. It begins at a fraction of 0.1 at log mass 9.9, dipping to a fraction of 0.08 at log mass 10.19 and gradually increasing from a fraction of 0.1 at log mass 10.46 to a fraction of 0.5 at log mass 11.4. The neither category rises from a fraction of 0.1 at log mass 9.75, peaking at a fraction of 0.19 at log mass 10.72 and then dipping to a fraction of 0.14 at log mass 11.04.
Summary: In all cases the log mass fraction value for the neither category does not exceed 0.2. For the merging category the fraction value generally exceeds a value of 0.4. At log mass 10.4, 10.5, the fraction for merging increases dramatically, whereas for the neither category this does occur, but is less pronounced and levels off or decreases.
Please excuse any glaring errors or typos.
Posted

by ChrisMolloy
Hereâ€™s the update of the QS 1149 with the QC 1196 asymmetrical merging and neither categories.
Asymmetrical are QC 58 merger, QC 123 neither; QS 141 merger, QS 158 neither. Zband estimated absolute magnitudes are 20.0 or brighter. Log Mass values for QS are mlpeckâ€™s values and categories are based on the labels consensus (threshold â‰¥ 0.5?) here (first and second posts in this thread). Six log mass bins are examined which are 010, 1010.3, 10.310.6, 10.610.9, 10.911.2, > 11.2. Both categories are examined in 4 redshift bins, z=0.020.04, z=0.040.06, z=0.060.08, z=0.080.10.
Error bars are Bayesian, and fractions are the Beta mean with a (0.5,0.5) prior. Where error bars are solid black lines this is the same fraction for both categories, e.g. last log mass bin in chart below.
What seems to be apparent, even though the fractions are quite small, is that for QC the neither category appears to be dominant at the lower log mass through to the higher log mass. The merger is less prevalent. For QS the opposite appears to be the norm, with the merger fraction starting of smaller than neither at the lower log mass and rising above neither around log mass 10.4 steadily increasing. The neither category is again wave like with higher fractions at the lower log mass than merger but declining or levelling off at the higher log mass.
In the charts that follow, the redshift cuts, this trend is generally replicated. Fractions are generally higher for QC neither at the lower and higher log mass (steadily declining), merger fractions are generally lower than neither and there is no significant increase at log mass 10.4 (one exception below). Whereas for QS the fractions for merging increase around log mas 10.4 and the neither category has a small rise around log mass 10.4 and then declines or levels off.
This chart is an exception. Similar in some senses to the QS charts, with the exception that QS and QC both rise together in the last bin rather than declining. Also, there is an error bar for the fourth neither bin but it is so small as to be nearly unnoticeable.
BTW. Previously I had been incorrectly calculating the log mass mean/median only for those galaxies that were asymmetrical in each bin. This is now amended so that the mean for the whole bin is calculated. It didnâ€™t affect the previous posts in any dramatic way.
Posted

by JeanTate in response to ChrisMolloy's comment.
Nice work, ChrisMolloy! ðŸ˜ƒ
I have a few questions about the data, and plots, which I'd like to cover before getting onto interpretations (hope you don't mind).

'neither' is '!merger' (not 'merger'), right?

You write "are based on the labels consensus (threshold â‰¥ 0.5?)"; what's the question mark for?

Is the length of each error bar "Â±1 sigma" (i.e. the interval [0.16,0.84])?

The 1149 QS and 1196 QC objects are each divided into four classes, using two independent criteria. Using the shorthands 'A' (asymmetrical), '!A' (not asymmetrical), 'M' (merger) and '!M' (neither), are the numbers in each of the four classes as follows (sorry, I can't do tables ðŸ˜¦):
QS: 141 A&M, 158 A&!M, ? !A&M, ? !A&!M
QC: 058 A&M, 123 A&!M, ? !A&M, ? !A&!M
The values of 'mean log mass' (xaxis), for each bin: there are two, one for 'merger' and one for 'neither', right? Would you mind picking a bin  the lowest mass bin ('010') for 'asymmetrical QS z=0.080.1', say  and giving the number of objects and the means?

"Also, there is an error bar for the fourth neither bin but it is so small as to be nearly unnoticeable."  smaller than the size of symbol used to plot the datapoint? the thickness of the combined lines? That would make it the smallest error bar in all the plots, right?

Suggestion: use the same yaxis scale for all plots (it's hard to compare one that is [0,1] with one that is [0,0.6], for example; can wait until last of course)
Posted


by ChrisMolloy in response to JeanTate's comment.
Thanks for the questions. I was a bit scant on details.
neither' is '!merger' (not 'merger'), right?
Correct, asymmetrical not merging.
You write "are based on the labels consensus (threshold â‰¥ 0.5?)"; what's the question mark for?
Should have expanded this. I wondered seeing as the spreadsheets were labelled consensus whether decisions were made on borderline classifications. More thinking aloud. e.g. 0.45 moving to a 0.5, if a galaxy clearly fitted a 0.5 classification by observation and external data. Not just the classification.
Is the length of each error bar "Â±1 sigma" (i.e. the interval [0.16,0.84])?
Correct. Should have noted this. My apologies.
The 1149 QS and 1196 QC objects are each divided into four classes, using two independent criteria. Using the shorthands 'A' (asymmetrical), '!A' (not asymmetrical), 'M' (merger) and '!M' (neither), are the numbers in each of the four classes as follows (sorry, I can't do tables ðŸ˜¦):
QS: 141 A&M, 158 A&!M, ? !A&M, ? !A&!M
QC: 058 A&M, 123 A&!M, ? !A&M, ? !A&!MCorrect. Will use correct notations in future. I think you might have mentioned this to me before. My apologies.
The values of 'mean log mass' (xaxis), for each bin: there are two, one for 'merger' and one for 'neither', right? Would you mind picking a bin  the lowest mass bin ('010') for 'asymmetrical QS z=0.080.1', say  and giving the number of objects and the means?
Here it is. QS z=0.080.1, log mass bin ('010') A&!M n 29, f 5, prior Beta( 0.5 0.5) mean 0.183333333, Â±0.06917690; A&M n29, f2, prior Beta (0.5,0.5), mean 0.083333333, Â±0.04757345.
Also, there is an error bar for the fourth neither bin but it is so small as to be nearly unnoticeable."  smaller than the size of symbol used to plot the datapoint? the thickness of the combined lines? That would make it the smallest error bar in all the plots, right?
Correct. The mean for that bin is 0.009615385, Â±0.009339241. The top bar is just visible. Will use smaller dataponts than what I am using in future. Thicknes of the combined lines  when I plotted this in excel only the neither line came out. So I increased the width of the merger line to exclude this (making it thicker and more blurred in colour) as it should show both neither and merger. Don't know why this happened. Possibly because the fractions for the last two bins are equal and neither takes precedent or maybe the colours I'm using?
Suggestion: use the same yaxis scale for all plots (it's hard to compare one that is [0,1] with one that is [0,0.6], for example; can wait until last of course)
I wondered about doing that. Initially, I just wanted to show equal comparisons between each bin.
Posted

by JeanTate
Thanks Chris.
One more try on the tables ...
[me] The 1149 QS and 1196 QC objects are each divided into four classes, using two independent criteria. Using the shorthands 'A' (asymmetrical), '!A' (not asymmetrical), 'M' (merger) and '!M' (neither), are the numbers in each of the four classes as follows (sorry, I can't do tables ðŸ˜¦):
It seems that the 'code' icon, when applied to text that contains symbols like ! and ?, refuses to format correctly. Here's sorta what I meant, a 2x4 table with headers:
SC: AandM AandN SandM SandN QS: 141 158 x1 y1 QC: 58 123 x2 y2
"AandM" here means "asymmetric and merger", "N" is "neither" (i.e. not merger), and "S" "symmetric" (i.e. not asymmetric). I was trying to ask this: what are x1, x2, y1, and y2?
[you] Correct. Will use correct notations in future. I think you might have mentioned this to me before. My apologies.
It isn't so much the notation  I was simply trying to be terse  as clarity, in order to create a simple table. But I got tripped up by Talk's (unwritten!) rules on formatting. Hope that's clearer now.
[you] Here it is. QS z=0.080.1, log mass bin ('010') A&!M n 29, f 5, prior Beta( 0.5 0.5) mean 0.183333333, Â±0.06917690; A&M n29, f2, prior Beta (0.5,0.5), mean 0.083333333, Â±0.04757345.
Thanks. So what is the mean log_mass for the 34 (=29+5) QS objects, and similarly the mean log_mass for the 31 (=29+2) QC ones?
On to interpretations ... ðŸ˜ƒ
General comment: if error bars overlap, the likelihood that the two datapoints are statistically different is relatively small (I'm sure there's a more rigorous way to say this, using words like "null hypothesis cannot be rejected" ...). Similarly, if a straight line goes through all relevant error bars, then that's statistical evidence for the existence of a linear trend (maybe mlpeck or a SCIENTIST could chime in here?)
For example, in the "Asymmetrical QC z=0.020.04" plot, all four pairs of error bars overlap, so the "neither" and "merger" fractions are the same, in each of the four bins. Further, you could draw a straight line y=0.12 (I'm guessing the number, judging by eye) and it would go through all eight error bars (so the data are consistent with the fractions being constant/not varying with mass).
On, then, to what I see in these plots (judging 'by eye'; numerical analyses needed!) ...
In the QS plots, the 'merger' fractions (the fractions of 'merger' objects which are asymmetric) rise as mass increases, for all z ranges, monotonically (or close to it). However, the 'neither' fractions do not show any obvious trend with mass (there's 'waviness', but it's not consistent, and a horizontal line at ~0.15 would fit most plots pretty well, even the overview one).
In the QC plots, the 'merger' and 'neither' datapoints are mostly the same (the z=0.060.08 plot is the main exception), and a gently sloping downward linear trend likely fits most too. However, a somewhat better fit might be a horizontal 'merger' line, and a downtrending line for 'neither'.
Comparing the QS and QC plots: the 'neither' data are ~the same, in most mass bins, in all (?) plots. However, the 'merger' data are anything but ~the same! ðŸ˜®
If the above holds up to further analysis, this leads to a very interesting pair of conclusions; namely,
 that for 'neither' objects, the fraction which zooites perceived to be asymmetric is constant (~15%), across all mass ranges, and
independent of whether the objects are 'post quenched' or not  perceived asymmetric fraction for merging QS objects is a (linear?) function of log_mass; the greater the mass, the more QS galaxies seem asymmetric (relatively; or something like this).
Cool! ðŸ˜„
Posted
 that for 'neither' objects, the fraction which zooites perceived to be asymmetric is constant (~15%), across all mass ranges, and

by ChrisMolloy
Sorry for the delay in responding.
Here's the tables for the charts. Hope this will suffice for the log mass, numbers and fractions questions. My formulas never transferred over to Google spreadsheets. First time using it. Possibly will redo them in google and repost later. Hope they make sense.
QC 1196 QS 1149 Log Mass Totals
Here's the rest of the data for the merging, neither, symmetrical and asymmetrical categories for QS and QC. What seems apparent is that there is a difference between galaxies that are AandM (asymmetrical and merging) and SandM (symmetrical and merging). Asymmetrical galaxies have a greater number of mergers (probably an obvious observation) than symmetrical ones. The SandM (symmetrical and merging) numbers for QS and QC are roughly equal (generalisation) and less than the AandM number. But, the clear difference, is that asymmetrical QS galaxies have a far greater number of mergers than asymmetrical QC galaxies. QS AandM galaxies are over two times more likely to be involved in a merger than a QC AandM galaxy. Furthermore, Quench galaxies are more likely to be asymmetrical than the Control ones. That is quite obvious looking at the four different categories and the numbers, the AandN and SandN columns are good examples. More asymmetrical than symmetrical.
Asymmetrical neither galaxies, in particular QC AandN is the biggest surprise. I think I would concur with your view
that for 'neither' objects, the fraction which zooites perceived to be asymmetric is constant (~15%), across all mass ranges, and independent of whether the objects are 'post quenched' or not
Regarding your comment here:
[you] Here it is. QS z=0.080.1, log mass bin ('010') A&!M n 29, f 5, prior Beta( 0.5 0.5) mean 0.183333333, Â±0.06917690; A&M n29, f2, prior Beta (0.5,0.5), mean 0.083333333, Â±0.04757345.
Thanks. So what is the mean log_mass for the 34 (=29+5) QS objects, and similarly the mean log_mass for the 31 (=29+2) QC ones?I omitted to put the prior Beta value in here. Should have been A&!M n30, f5.5, A&M n30, f2.5. Hope the tables will suffice.
Will comment on interpretations more later, but initially they do seem good.
Posted

by JeanTate in response to ChrisMolloy's comment.
Thanks! ðŸ˜ƒ
I'll check this out in detail, later, but it looks good (quickly eyeballing it).
This is, I think, something which Tools was intended to be able to provide. The good news is that we (well, actually you) were able to find a workaround ... ðŸ˜‰
Assuming we reach agreement on interpretations, in the next few days, we will have gone about as far as we can, without further input from one or scientist, wouldn't you say?
Posted

by ChrisMolloy in response to JeanTate's comment.
Sorry for the delay in responding. Here is a more detailed analyses of the log mass data. I think the conclusions I've drawn here basically concur with what you've written in the above posts. I've added extra decimal places in some areas to highlight the slight differences between the fractions. Also I've added a qualifier when examining the attached spreadsheets above, in particular columns c and d at the bottom of this post.
Log Mass Overview
The AandN (asymmetrical and neither) category for QC asymmetrical begins with a fraction of 0.18 at log mass 9.81, drops to a fraction of 0.1 at log mass 10.16 and then gradually declines, with a fraction of 0.09 at log mass 10.43, a fraction of 0.0698 at log mass 10.72, and a fraction of 0.0692 at log mass 10.98. There is then a slight rise to a fraction of 0.07 at log mass 11.26. With the exception of log mass 9.81, the fractions are equal to and less than 0.1 for those bins. QC AandM (asymmetrical and merger) the fractions begin at log mass 9.81 with a fraction of 0.07, dip to a fraction of 0.05 at log mass 10.16 and a fraction of 0.03 at log mass 10.43. The fractions then rise to a fraction of 0.06 at log mass 10.72, dip again to a fraction of 0.03 at log mass 10.98, and then rise to a fraction of 0.07 at log mass 11.26. The fractions for all of the six log mass bins are less than 0.1 and less than AandN, with the exception of log mass 11.26 where the fractions are equal to AandN.QS Asymmetrical log mass overview generally reflects previous posts. Fractions for AandN begin at a fraction of 0.15 log mass 9.83, drop to a fraction of 0.11 at log mass 10.16, rise to a fraction of 0.12 at log mass 10.43 and a fraction of 0.20 at log mass 10.72, and then decline to a fraction of 0.11 at log mass 10.99 and a fraction of 0.1 at log mass 11.31. For AandN the fractions donâ€™t exceed 0.2 and are between the range of 0.1 and 0.2. For AandM the fraction rise is completely monotonic, starting at a fraction of 0.06 at log mass 9.83, a fraction of 0.07 at log mass 10.16, a fraction of 0.08 at log mass 10.43, a fraction of 0.25 at log mass 10.72, a fraction of 0.39 at log mass 10.99, and rising to a fraction of 0.5 at log mass 11.31. There is a sharp increase in fractions at log mass 10.43 onwards for both AandN and AandM, with AandM monotonically rising and AandN declining from log mass 10.72 onwards.
Redshift z=0.020.04  QC AandN begins with a fraction of 0.21 at log mass 9.75, declines to fraction of 0.08 at log mass 10.16, rises to fraction of 0.20 at log mass 10.40, and then declines to a fraction of 0.13 at log mass 10.72. QC AandM begins with a fraction of 0.07 at log mass 9.75, rises to a fraction of 0.08 at log mass 10.16 and a fraction of 0.12 at log mass 10.40, and then drops to a fraction of 0.04 at log mass 10.72. Fractions for AandM are lower than those of AandN, slightly increasing with increased log mass and then declining. AandN fractions are more of a declining wave with increased fractions at the lower log mass, and lower at the higher log mass.
QS AandN begins at a log mass of 9.79 with a fraction of 0.05, rises to a fraction of 0.17 at log mass 10.15, dips to a fraction of 0.10 at log mass 10.42, rises to a fraction of 0.25 at log mass 10.74 and then declines to a fraction of 0.16 at log mass 10.92. QS AandM begin with a fraction of 0.018 at log mass 9.79, dips to a fraction of 0.015 at log mass 10.15 and then gradually increases, with a fraction of 0.17 at log mass 10.42, a fraction of 0.25 at log mass 10.74 and a fraction of 0.83 at log mass 10.92. AandN fractions are slightly higher than AandM at the lower log mass but do show a gradual increase, again a wave like function, from the lower log mass to the higher log mass. AandM galaxies with the exception of a slight decline at the lower log mass exhibit a fraction increase with increasing log mass from log mass 10.15 onwards, surpassing the AandN fractions.
Redshift z=0.040.06 â€“ QC AandN begins at log mass 9.80 with a fraction of 0.14, dips to a fraction of 0.08 at log mass 10.15, rises to a fraction of 0.09 at log mass 10.42, dips to a fraction of 0.04 at log mass 10.73 and then rises again to a fraction of 0.06 a log mass 10.99. A general observation is that this reflects a shallow decreasing wave or step function. QC AandM begins at log mass 9.80 with a fraction of 0.07, increases to a fraction of 0.08 at log mass 10.15 and then declines to a fraction of 0.01 at log mass 10.42. The fractions then increase to a fraction of 0.04 at log mass 10.73 and a fraction of 0.06 at log mass 10.99. QC AandM general observations are that the fractions slightly rise then decline sharply and then slightly rise again, but at a lesser fraction level than it initially rose to. Again, this could be categorised as a decreasing wave or step function. With the exception of AandN log mass 9.80, fraction 0.14, all fractions for both categories are less than 0.1.
QS AandN begins at a fraction of 0.19 at log mass 9.81 and dips to a fraction of 0.05 at log mass 10.15. The fractions then increase to a fraction of 0.11 at log mass 10.41, and a fraction of 0.27 at log mass 10.72, before declining to a fraction of 0.07 at log mass 10.95. In terms of general observations AandN fractions are higher at the lower log mass before dipping, and then there is a significant rise from log mass 10.41 onwards before declining at log mass 10.72. QS AandM begins with a fraction of 0.04 at log mass 9.81, a fraction of 0.06 at log mass 10.15, a fraction of 0.11 at log mass 10.41, a fraction of 0.32 at log mass 10.72 and a fraction of 0.5 at log mass 10.95. QS AandM the fraction rise is completely monotonic. The fractions begin lower than AandN but surpass this category from log mass 10.15 onwards.
Redshift z=0.060.08. QC AandN begins with a fraction of 0.18 at log mass 9.83, declining to a fraction of 0.11 at log mass 10.16, a fraction of 0.07 at log mass 10.42, and a fraction of 0.009 at log mass 10.71, before rising sharply to a fraction of 0.06 at log mass 10.98 and a fraction of 0.25 at log mass 11.35. In terms of general observations this is the exception to previous redshift cuts, with a sharp rise at the higher log mass of 10.71 onwards, rather than a gradual decline, which is present up until log mass 10.71. QC AandM begins with a fraction of 0.05 at log mass 9.83, dips to a fraction of 0.02 at log mass 10.16, rises to a fraction of 0.06 at log mass 10.42, dips to a fraction of 0.04 at log mass 10.71 and rises to a fraction of 0.06 at log mass 10.98 and a fraction of 0.25 at log mass 11.24. General AandM observations are that the fractions are almost even in their wave like appearance but with a sharp rise from log mass 10.98 onwards.
QS AandN again begins with higher fractions at the lower log mass and gradually declining fractions with higher log mass. Beginning with a fraction of 0.17 at log mass 9.84 dropping to a fraction of 0.12 at log mass 10.17, a fraction of 0.08 at log mass 10.42 and then rising to a fraction of 0.15 at log mass 10.70 before dropping to a fraction of 0.11 at log mass 11.00. QS AandM starts at a fraction of 0.10 at log mass 9.84, dipping to a fraction of 0.09 at log mass 10.17 and a fraction of 0.07 at log mass 10.42, before rising to a fraction of 0.19 at log mass 10.70 and a fraction of 0.38 at log mass 11.00. General observations for both categories is that at log mass 10.42 onwards the fractions for AandN and AandM both rise, with the former dropping away at log mass 10.70 but with the latter steadily increasing with increased log mass. Again AandN has a higher fraction at the lower log mass.
Redshift z=0.080.1 QC AandN begins again at a higher log mass than AandM, with a fraction of 0.31 at log mass 9.87. It then declines to a fraction of 0.10215 at log mass 10.18, rising slightly to a fraction of 0.10273 at log mass 10.44 and a fraction of 0.10747 at log mass 10.72 before declining to a fraction of 0.1 at log mass 10.98 and a fraction of 0.08 at log mass 11.24. Fractions here, excusing the increased decimal places to highlight the differences, are essentially horizontal with the exception of the higher fraction 0.31 at log mass 9.87 and the lower fraction of 0.08 at log mass 11.24. QC AandM the fractions begin with a fraction of 0.16 at log mass 9.87 and drop to a fraction of 0.09 at log mass 10.18 and a fraction of 0.02 at log mass 10.44. The fractions then rise to a fraction of 0.08 at log mass 10.72, dipping to a fraction of 0.04 at log mass 10.98 and then rising to a fraction of 0.08 at log mass 11.24. QC AandM fractions are generally again lower than AandN with the exception of log mass 11.24 where they are both equal. With the exception of log mass 9.87 fraction 0.16 all fractions for AandM are less than 0.1.
QS AandN fractions begin with a fraction of 0.18 at log mass 9.88, drop to a fraction of 0.15 at log mass 10.16, rise to a fraction of 0.16 at log mass 10.44 and a fraction of 0.22 at log mass 10.73 before declining to a fraction of 0.12 at log mass 10.99 and a fraction of 0.1 at log mass 11.31. QS AandM fractions again begin lower than AndN, starting at a fraction of 0.083 at log mass 9.88, rising to a fraction of 0.088 at log mass 10.16, and then dipping to a fraction of 0.083 again, at log mass 10.44. The fractions then rise steadily, a fraction of 0.27 at log mass 10.73, a fraction of 0.34 at log mass 10.99, and a fraction of 0.5 at log mass 11.31. Again for both AandN and AandM there is a rise in fractions at log mass 10.44 with the AandN fraction dropping away at log mass 10.73 onwards, and the AandM fraction steadily rising.
Error Bars In terms of error bars and the significance of the data presented, most fraction error bars do seem to overlap or come near to overlapping for at least over half of the data (bins) presented in each diagram. From what little Iâ€™ve read on overlapping error bars I canâ€™t offer an opinion either way on the significance of the data presented. QS there seems to be less overlap at the lower log mass and the higher log mass with overlap occurring in between. For QC the overlapping error bars are more scattered throughout the data.
Spreadsheets Also, when looking at the attached spreadsheets, each table is for the individual category, either QS AandN or AandM and QC AandN or AandM. Columns C and D reflect the actual total number of galaxies for each bin. The column title A is those asymmetrical, and NA is for those not asymmetrical. So for AandM (asymmetrical and merging) A stands for the number of galaxies asymmetrical and merging, and NA is the number of galaxies not asymmetrical and merging. For AandN (asymmetrical and neither) A stands for the number of galaxies asymmetrical and neither and NA stands for the number of galaxies not asymmetrical and neither. There are, therefore, three possible variables in the NA field. For AandM the variables could be AandN, SandM (symmetrical and merging), and SandN (symmetrical and neither) and for AandN the variables could be AandM, SandN and SandM. I hope this didnâ€™t cause any ambiguity.
Posted