Subjectively deterministic-looking features of scatter plots

  • Level: Graduate 
  • Thread starter Swamp Thing
  • Start date
  • Tags Tags
    Plots
2 replies · 2K views · last reply
Swamp Thing
Insights Author
Messages
1,080
Reaction score
851
Scatter plot of many black dots spread across a horizontal range from about -0.5 to 1.5 and vertical range from -4 to 4, dense near the origin and widening to the right.


This is a scatter plot of the points where ##|\zeta(1/2 + it)|## has local maxima.

On seeing it, my first thought was that there seems to be a certain amount of deterministic quality, a sort of "texture", in the scatter. There seem to be groups of points that look like skeins, closed curves etc. There are white spaces roughly bounded by closed curves, and "no-go" areas surrounded by dense populations. It is as if some kind of "attractor" phenomenon is at work underlying the randomness.

So as an example of an "attractor" modulating a chaotic scatter, here are two plots of the so-called Hopalong Attractor discovered by Barry Martin and popularized in Scientific American a few decades ago.

The first one is a zoomed-in version of the second one.
Scatter plot of many small black dots spread fairly evenly over a square region, with horizontal axis from about 1.3 to 2.3 and vertical axis from 0 to about 2.8.
Scatter plot of dense black points on x-y axes spanning about -5 to 5, forming a symmetric star-like pattern with curved arms and white gaps, labeled ticks 2, 4 and -2, -4.


But there was always the possibility that the features I perceived in the first plot were all in my head, so I plotted a Gaussian distribution and zoomed in on the middle, just to see if the same apparent groupings appeared:

Scatter plot of many small black dots on white, spanning x and y from -2 to 2 with axes marked; dots are denser near the centre and show loose clustering and gaps.

To my eye, this seems to have a surprising amount of "texture", with groups of points seeming to form somewhat non-random curves, skeins, and even bounded white spaces here and there. This suggests that any deterministic appearance I see in the first plot is entirely subjective, i.e. "all in my head".

But I can't quite make up my mind. After all, it is not unreasonable that the peaks of a quasi-periodic function could show some pattern underneath the randomness. And the first plot (peaks of zeta) has a qualitatively different and more distinct texture.

So... any thoughts?

Also wondering if there have been studies of the psychology of pattern discernment in random or nearly random distributions.
 
Swamp Thing said:
On seeing it, my first thought was that there seems to be a certain amount of deterministic quality, a sort of "texture", in the scatter. There seem to be groups of points that look like skeins, closed curves etc. There are white spaces roughly bounded by closed curves, and "no-go" areas surrounded by dense populations. It is as if some kind of "attractor" phenomenon is at work underlying the randomness.
I don't see any of that.

Swamp Thing said:
To my eye, this seems to have a surprising amount of "texture", with groups of points seeming to form somewhat non-random curves, skeins, and even bounded white spaces here and there.
I also don't see any of that.

Swamp Thing said:
Also wondering if there have been studies of the psychology of pattern discernment in random or nearly random distributions.
I am not sure that this is related, but there are studies that show that it is very difficult for humans to actually produce a random distribution. We tend to produce distributions of points that are much more uniform than is correct for a truly random distribution. If it is random then there will randomly be clumps and voids. A distribution that avoids any clumps and voids would be uniform, not random.
 
Swamp Thing said:
Also wondering if there have been studies of the psychology of pattern discernment in random or nearly random distributions.
Different people will believe they see different things in a random distribution. That is because the brain tries to extract features to make sense of the pattern. Different people develop different neural networks because they have been subjected to different learning experiences.