Big data and precision
本文指出大数据结构复杂，标准统计方法得出的精度估计可能误导，即使参数点估计尚可。作者提出一种误差累积的通用表示法，探讨其对回归系数估计的影响，并联系时间序列中短期与长期依赖的区别。
So-called big data are likely to have complex structure, in particular implying that estimates of precision obtained by applying standard statistical procedures are likely to be misleading, even if the point estimates of parameters themselves may be reasonably satisfactory. While this possibility is best explored in the context of each special case, here we outline a fairly general representation of the accretion of error in large systems and explore the possible implications for the estimation of regression coefficients. The discussion raises issues broadly parallel to the distinction between short-range and long-range dependence in time series theory.