# PREPRINT 13/1999

Alexander GNEDIN and Sergei KEROV

## A CHARACTERIZATION OF GEM DISTRIBUTIONS

This preprint was accepted September 11, 1999
Contact: S. V. Kerov

ABSTRACT:
Let $(X_n)$ be a residual allocation model with
i.i.d. residual fractions $U_n$. For a random variable $W$ with
values in $[0,1]$ independent of $(X_n)$ we define another
sequence $(Y_n)$ by setting
(Y_1,Y_2,Y_3,Y_4,\ldots) = \left\{ \aligned &(W,X_1-W ,X_2,X_3,\ldots)\text{ \ \ \ if \ } W < X_1, \\ &(X_1+X_2,X_3,X_4,X_5,\ldots) \text{ \ if \ } W \ge X_1. \endaligned\right.
Under minor regularity assumptions we show that $(X_n)$ and
$(Y_n)$ have the same probability law if and only if this law is
a GEM distribution. In this case, the distribution of $W$ and the
$U_n$'s is beta $(1,\theta)$ for some $\theta>0$

[ Full text: (.ps.gz)]
Back to all preprints
Back to the Petersburg Department of Steklov Institute of Mathematics
