1: \begin{abstract}

2: We consider parameter estimation, hypothesis testing and variable

3: selection for partially time-varying coefficient models. Our asymptotic

4: theory has the useful feature that it can allow dependent,

5: nonstationary error and covariate processes. With a two-stage method,

6: the parametric component can be estimated with a $n^{1/2}$-convergence

7: rate. A simulation-assisted hypothesis testing procedure is proposed

8: for testing significance and parameter constancy. We further propose an

9: information criterion that can consistently \mbox{select} the true set of

10: significant predictors. Our method is applied to autoregressive models

11: with time-varying coefficients. Simulation results and a real data

12: application are provided.

13: \end{abstract}