Finding the perimeter points of an area of points.

[brophymj]

I have a dataset of a 100 randomly generated xy coordinates. I'm trying isolate the points that are on the edge of the area. The definition of this would be:

if the xy coordinate has no x value from another point above it or

if the xy coordinate has no x value from another point below it or

if the xy coordinate has no y value from another point above it or

if the xy coordinate has no y value from another point above it

I'm looking for a reasonably efficient way of doing this.,

data obs;

   do i= 1 to 100;

      x= round(1000* ranuni(123),1);

      y= round(1000* ranuni(789),1);

      output;

   end;

run;

[ballardw]

I think you may need to provide some graphic examples of what you mean by coordinates that are on the "edge".

Same x but different y, or same y but different x seems to be the implied condition. But why would that make them an "edge"

If you have

(0, 10)

(0,15)

(0,20)

( 5, 10)

( 5, 15)

(5,20)

(10,10)

(15,10)

(25,10)

which are "edge"?

[brophymj]

All except (5,15). When I say edge I mean if you drew a line arounf all the outer points such that there were no points outside the shape the points joining the line would be on the edge. Like an efficient frontier.

[Rick_SAS]

It sounds like you are looking for the convex hull of the set of planar points. There is a convex hull algorithm in SAS/IML:

SAS/IML(R) 14.1 User's Guide