Optimal Design for Estimation of Variance in Nonparametric Regression Using First Order Differences
研究了非参数回归中基于一阶差分的方差估计量的最优设计，发现均匀设计在多数情况下近似最优，并讨论了最差设计。
In nonparametric regression, the variance of the response can be estimated by the sum of squares of differences of the observed response. In this paper we obtain the most efficient design for a general variance estimator defined by first order differences. It is found that for this estimator, in the majority of cases, a good approximation to the most efficient design is the uniform design. The least efficient designs are also discussed.