Statistica Sinica 29 (2019), 1963-1975
Abstract: Failure time data occur in many fields and in various forms and, as a result, have been studied extensively by researchers. To handle such data, many semiparametric regression models have been proposed and investigated, including the proportional hazards model and the additive hazards model. In this study, we consider a class of partially linear varying-coefficient transformation models. Given the limitations of existing inference procedures, we propose a more general rank estimation procedure. Furthermore, we establish the finite and asymptotic properties of the resulting estimators of the regression parameters and illustrate the proposed procedure by means of an example.
Key words and phrases: B-splines, smoothing partial rank, transform model, varying coefficients, weighted bootstrap.