A method for the numerical inversion of Laplace transform on the real line
of a heavy-tailed probability density function is presented. The method ass
umes as known a finite set of values of the Laplace transform and chooses t
he analytical form of the approximant maximizing Shannon-entropy, so that p
ositivity of the approximant itself is guaranteed. The problem resorts to a
finite generalized moment problem and a stability analysis is also provide
d. Some numerical results are illustrated. (C) 2001 Elsevier Science Inc. A
ll rights reserved.