Jump-preserving varying-coefficient models for nonlinear time series

Abstract An important and widely used class of semiparametric models is formed by the varying-coefficient models. Although varying coefficients are traditionally assumed to be smooth functions, model considered here with coefficient functions containing a finite set discontinuities. Contrary existing nonparametric estimation piecewise under dependence applicable in time series heteroskedastic serially correlated errors. Additionally, conditional error variance allowed exhibit discontinuities at points too. The (uniform) consistency asymptotic normality proposed estimators established finite-sample performance tested via simulation study real-data example.