计算机模拟正态随机数的产生与统计检验

Generation and Statistical Test of Computer Simulated Normal Random Numbers

提出一种产生正态随机数的计算机新算法。这种新算法是在现有的乘同余法和混 合同余法等常用算法的基础上改进而来。统计检验表明：用原常用算法产生的随机数， 例如先用乘同余法或混合同余法产生均匀分布随机数，再用抽样变换法构成正态随机 数，其分布特性令人满意，但独立性质量不高；而用新算法产生的随机数，既能保持 原来分布特性较好的特点，又在独立性上有较大的改进。

A new computer algorithm of generating normal distributed random numbers is presented. The new algorithm is obtained by improving the multiplicative congruential and the mixed congruential methods which are in common use now. Statistical test results show that, for the random numbers generated by the methods in common use, such as the normal random numbers constructed by sample transformation method from the uniform random numbers generated by the multiplicative congruential algorithm or by the mixed congruential algorithm, the distribution performance is satisfactory, but the independence performance is quite unsatisfactory, and for the random numbers generated by the new algorithm, not only the good distribution is maintained but also the independence is improved considerably.