Generating Random Sampling Numbers

Abstract

Simulation experiments with the help of random numbers are increasing in various Defence application areas. Some authors have brought out autocorrelations in the sequences generated by the computer methods. To overcome this defect, a new algorithm based on the concept of random sampling (with replacement) from a finite population has been developed. The constants in the linear congruential relation are chosen so that (a) the proportion of distinct numbers in the generation compares with the theoretical expectation and (b) the period of the generator covers almost the edtire population of 16-bit numbers. The criteria suggest to choose the variable multiplier (ai =2ai-1 +' 1) with initial value a. = 2^9+ 5 or 2^10 + 3. The randomness of the generation has been compared with that of a Bernoullian sequence.