An upper bound for the correlation ratio of records

We obtain an upper bound for a measure of the performance of the least squares predictor of the jth record of a sequence of continuous i.i.d random variables as a function of the ith record. We show also that such bound is attainable, except for location and scale parameters, by exponential distributions. This work was supported in part by Ministerio de Educación y Ciencia through Grants PB96-1416-CO2-02 and HA 1997-0123.