zbMATH — the first resource for mathematics

Robust regression with censored and truncated data. (English) Zbl 0941.62080
Ghosh, Subir (ed.), Multivariate analysis, design of experiments, and survey sampling. A tribute to Jagdish N. Srivastava. New York, NY: Marcel Dekker. Stat., Textb. Monogr. 159, 231-263 (1999).
From the paper: This paper gives a unified exposition of recent work on \(R\)- and \(M\)-estimators of regression parameters based on left-truncated and right-censored (l.t.r.c.) data. It also introduces a relatively simple algorithm to compute \(M\)-estimators. Starting with a good preliminary estimate described in Section 2.4, the algorithm typically converges after a few Gauss-Newton-type iterations. A more extensive search involving the simplex method is used to supplement these iterations in the extraordinary situation when they have not converged before a prespecified upper bound on the number of iterations. This computational method makes \(M\)-estimators much more attractive than \(R\)-estimators, which have similar robustness properties but much higher computational complexity. It also makes the computationally intensive bootstrap methods in Section 2.6 and the “leave-one-out” regression diagnostics in Section 3 feasible for the \(M\)-estimators based on l.t.r.c. data. For multivariate covariates, graphical methods for displaying the censored and truncated residuals are given in Section 3 to assess the adequacy of the regression model.
For the entire collection see [Zbl 0927.00053].
62J05 Linear regression; mixed models
62F35 Robustness and adaptive procedures (parametric inference)
62G08 Nonparametric regression and quantile regression
62J20 Diagnostics, and linear inference and regression
65C60 Computational problems in statistics (MSC2010)