Statistical tests for clustering of second phases in composite materials

Abstract
A statistical comparison is made of various tests for the randomness or otherwise of second-phase particles in composite materials. Quadrat, nearest-neighbour, Ripley's K function and Dirichlet tessellation-based tests are compared. It is found that a test based on Ripley's K function performs well in detecting both clustered and regular alternatives to randomness. Additionally these tests have the benefit of providing a suitable alternative model when randomness is rejected.