无论是在统计推断及模拟，还是科学计算中，对复杂的目标函数进行高效、准确的模拟抽样都是非常重要的一环。当面对的目标函数非常复杂，或其密度函数不完全已知时，经典的模拟抽样方法通常失效。当前使用最为广泛的是SIR和MCMC算法。但在很多情形下，SIR算法可能由于无法找到高效的提案分布而使得抽样效率较低；另外由于算法自身的马尔科夫性，收敛状态的检验、预热时间的确定等问题一直困扰着MCMC的使用。因此，找到一种简单高效，在前述SIR和MCMC方法效率低时仍能适用的算法。无论对统计推断，还是科学计算的发展都具有重要的意义。本项目拟借鉴空间填充设计中的均匀设计思想。将目标分布支撑进行有效离散化，将连续分布抽样问题转化为离散分布抽样，从而提出一种新的统计模拟抽样器。本项目的研究内容综合了统计模拟、试验设计、变量选择及Bootstrap等多个研究领域的理论和方法。对统计学的发展具有一定的理论和实际意义。

Generating samples from a given probability density function is an important issue in both statistics and scientific computing. In the literature, many methods were proposed for sampling unnormalized distribution, such as SIR and MCMC algorithms. An efficiency proposal function is necessary for SIR. And for MCMC, there are still exist some tricky problems to handle, such as how to testing the convergence of Markov Chain, and how long the burn-in time should be last. Under this grant support, we will propose a new sampling method, which uses a uniformly scattered quasi-random number set as candidate sampling pool. And the new sampling algorithm is a simple generating process. It can be easily implemented in to the common statistics program for any complex continuous target distribution. In this study, some related problems, such as statistical simulation, experimental design, variable selection, and bootstrap resampling, will also be concerned.