It's a difficult problem even in CAD.
The problem with a cloud of points in 3D space is that computer screens and ordinary printers don't give you depth cues, so you don't know how far a point is behind or in front of the viewing plane. And rotating the viewpoint just makes different sets of points appear to coincide.
[ Just last month I found two lines that behaved oddly with object snaps. One of them was ~14,000 miles behind the viewing plane. Go ahead; zoom extents and fix that.
No, I have no idea how it got that way, but I have indications that it may have been copied to hundreds of other drawings.
Found other lines at similarly bizarre and not clearly related distances, too. All were inside blocks nested in blocks that were used everywhere. It's the sort of thing you might leave behind if you were leaving on bad terms... the AutoCAD equivalent of a programmer's code bomb. ]
In a CAD system, you might be able represent 3D scatter data as a perspective view of a 2D array of two- element lines. I.e. a given data point would be represented by a faint black line whose length is proportional to the distance of a colinear colored line from the xy plane, and the length of the colored line would represent the range of z values for all points with the line's xy value. Extend that to use two colors if there are three points, etc. Of course there would be no colored line, or maybe just a colored dot, at the end of the black line if there's only one z value for that xy pair.
Mike Halloran
Pembroke Pines, FL, USA