Statistical Foundations of Data Science.

Finally, in the last section of this chapter, the authors describe data swap and gradient approximation to address asymptotic normal estimation and efficient inference in the semi-LD approach. Moreover, each chapter has exercise problems, which are crucial issues in the authors' research work and, at the same time, help readers review the materials in each chapter. Different from the PLS in Chapter 7, the authors clearly introduce the structure of GGM and summarize several useful methods, such as penalized likelihood, a.k.a. graphical lasso, and the constrained l SB 1 sb -minimization for inverse matrix estimation. Chapter 6 discusses M-estimation with penalization, which is regarded as the generalization of least squares estimations and likelihood estimations in Chapters 4 and 5, and M-estimation enables us to deal with a large body of different types of data structures. [Extracted from the article]