Comparative study for assessing the Pitman's closeness of predictors based on exponential record data

In this article, the performances of different point predictors of future record data based on some informative records from the two-parameter exponential distribution are studied in the sense of Pitman's measure of closeness. On the basis of one-sample and two-sample prediction problems, we apply the Pitman's measure of closeness to comparing the best unbiased, best invariant, maximum likelihood, conditional median predictors of the future records. Finally, all these predictors are compared with respect to the median unbiased predictor in the sense of Pitman closeness. Numerical computations for Pitman's closeness of all predictors are performed, presented and discussed to examine how these predictors are compared among themselves. On the whole, it is shown from our results that the best unbiased and conditional median predictors are very competitive predictors in terms of Pitman closeness when compared to all other predictors for both prediction problems.

Automated summary

The performances of different point predictors of future record data are studied based on informative records from the two-parameter exponential distribution, with a focus on Pitman's measure of closeness. The best unbiased and conditional median predictors are found to be competitive in terms of Pitman closeness when compared to other predictors.