MBS 94-05
Random Utility Representations of Finite m-ary Relations
Michel Regenwetter
Block and Marschak (1960) discussed the relationship between a probability
distribution over the class of strict linear rankings on a given set and
a family of jointly distributed random variables. The present paper generalizes
this random variable (random utility) representation to m-ary relations.
It specifies conditions on an N-dimensional random vector that allow us
to induce a probability distribution over a given collection of m-ary relations
on an N element set. Vice versa conditions are presented for a probability
distribution on such a collection of m-ary relations to induce an N-dimensional
random vector. In particular, random utility representations of betweenness
relations, weak orders and semi-orders are reported in examples. It is
shown that in these specific cases the conditions on the random vector
are either trivial or they reduce to the noncoincidence of the corresponding
family of random variables. Noncoincidence (Falmagne & Regenwetter,
1993) means that the probability of two random variables taking the same
value is zero. The paper thus extends the concept of random utility to
more general preference relations than strict linear orders.