Search

Chapter 8 simplifies the task to the estimation of the central mean subspace, replacing the SDR conditional independence assumption (that Graph HT ht for some matrix ) with an assumption on the conditional mean: Graph HT ht . To this point, the dimension of the SDR subspace has been taken as given, and Chapters 9 and 10 examine techniques for determining the dimension from data. Chapter 11 reexamines estimation of the central mean subspace with the outer product of gradients and minimum average variance estimators. [Extracted from the article]