A Reexamination of Finite- and Infinite-Variance Distributions as Models of Daily Stock Returns
比较了广义稳定帕累托分布与三种有限方差、时间无关分布对日股票收益序列的拟合效果，发现考虑偏度后有限方差模型仍占优，其中混合扩散跳跃和复合正态模型描述性最强。
Abstract This study investigates the general (asymmetric) stable Paretian distribution and three finite- variance, time-independent distributions applied to daily stock-return series. Previous empirical comparisons have, in general, ignored the existence and effects of skewness on the process parameters of stable laws. The results of log-likelihood ratio and log-odds tests indicate that finite-variance models still dominate after accounting for documented skewness. In particular, the mixed diffusion-jump and compound normal models appear to be the most descriptive time-independent models. KEY WORDS: Asymmetric stable Paretian distributionGoodness of fitSkewnessTime-independent processes