How to bound a set of data points from above?

Question

The following data produces a distribution as shown. However I want that black-line to be highlighted that is kind of bounding these points from above. How can one do this in Mathematica?

data=Uncompress[FromCharacterCode[
  Flatten[ImageData[Import["https://i.stack.imgur.com/FtJS8.png"],"Byte"]]]]

It depends what you mean by "upper bound". You might consider a quantile regression function written by @AntonAntonov: https://mathematica.stackexchange.com/questions/94770/upper-envelope-of-data/106173#106173. — JimB, Aug 09 '22 at 22:30

Bill · Accepted Answer · 2022-08-10T16:50:24.277

4

See if this does what you want:

data={...};
findslope[{x_,y_}]:=y/x;
allslopes=Map[findslope,data];
maxslope=Last[Sort[allslopes]];
Show[ListPlot[data],Plot[maxslope*x,{x,0,1}]]

Then you can fiddle with the display details to make your line bold and black

edited Aug 10 '22 at 16:50

answered Aug 09 '22 at 20:37

Bill

12,001
12
13

1

You could probably use Max[s] instead of Last[Sort[s]] – Lukas Lang Aug 09 '22 at 20:46
1

@LukasLang Agreed. For what seem to be very new users I try to write very simple very methodical code and do things in the smallest most deliberate steps. If they can understand it and use it and even "optimize" it for themselves then I hope they have learned more. More experienced users might have written all that in a single line consisting of at least half punctuation characters But they wouldn't have asked this. ;} – Bill Aug 09 '22 at 20:51
Thanks @Bill. That works! But how can I known the equation of that line? – seeker Aug 10 '22 at 07:49
1

@seeker From your data and from your graph it looks like your bound is y=max*x+0. That max*x was what I plotted in the code. I urge you to study each line of that code I wrote. Look up each of those function names in the help system. Look at all the examples. Try to figure out what I was thinking when I wrote each of those lines. There are several ideas and methods in that code that will really help you in the future if you can figure out how each of those works. I hope it works for you. – Bill Aug 10 '22 at 15:28
1

Thanks a ton!!! It really helped. – seeker Aug 10 '22 at 15:55

cvgmt · Answer 2 · 2022-08-12T01:34:59.300

Since the data not always looks contain the original {0,0},we consider to use ConvexHullMesh or ConvexHullRegion to get the flat boundary of data and use HighlightMesh to view the index of such line.

reg = ConvexHullMesh[data];
HighlightMesh[reg, {Style[2, Black], Style[1, Orange], 
  Style[0, Blue, AbsolutePointSize[6]], 
  Table[Labeled[{1, i}, 
    Style[i, White, FontFamily -> "Times", 20]], {i, 
    Length@MeshPrimitives[reg, 1]}]}, AspectRatio -> 1, 
 Epilog -> {Opacity[.5], Green, AbsolutePointSize[1.5], Point[data]}, 
 Background -> Gray]

7.

reg = ConvexHullMesh[data];
ListPlot[data, 
 Epilog -> {{Green, 
    MeshPrimitives[reg, Drop[MeshCellIndex[reg, 1], {7}]] /. 
     Line -> InfiniteLine, Red, 
    MeshPrimitives[reg, {1, 7}] /. Line -> InfiniteLine, 
    AbsolutePointSize[8], Point[{0, 0}]}}, PlotRange -> All]

How to bound a set of data points from above?

2 Answers2