I'm currently working on a Level of detail system for an XNA game, the system subdivides triangles with high screen space error and merges triangles with low screen space error.
World Space Error is a heuristic which estimates the error that this triangle has, so for example if there is a triangle which is on an almost flat surface, it'll have a very low screen space error because it's a very good approximation of that surface, however if there is a triangle on the surface of a sphere it'll have a higher screen space error because obviously splitting that triangle up into more triangles would get a better approximation of a spherical surface.
Screen Space Error is a slight modification of World Space Error, basically the error is modified so triangles closer to the camera, and closer to the centre of the field of view have a higher error score.
What is a good and efficient way to calculate screen space error?
Current Solution: Screen Space Error = World Space Error / dot([vector to triangle in question], [camera direction vector])