Two Practical Procedures for Estimating Multivariate Nonnormal Probability Density Functions
提出两种适用于小样本的多元非正态联合概率密度函数经验估计方法，一种通过边际分布变换为正态，另一种利用条件分布与边际分布的乘积，并回顾了多元正态性检验。
Abstract This article presents two procedures for empirically estimating nonnormal joint probability density functions (pdf's) that are operational with small samples. One procedure empirically estimates marginal distributions. Estimated marginal distributions are then used to transform variates to univariate normality; the transformed variates are assumed to have a multivariate normal distribution. The second approach exploits the identity that a joint distribution is the product of a conditional pdf and a marginal pdf. Conditional and marginal pdfs are individually estimated with this approach. Statistical tests for multivariate normality are also reviewed.