Statistical inference for autoregressive models under heteroscedasticity of unknown form (Q2284370)

Consider an ordinary \(p\)th order heteroscedastic autoregressive model (AR). Within the model, \(y_t\) is allowed to be nonstationary with a finite variance due to the heteroscedasticity of \(\epsilon_t\). First, the asymptotic normality of the weighted lead absolute deviations (LADE) for the model is established. The random weighting (RW) method is developed to estimate its asymptotic covariance matrix, leading to the implementation of the Wald test. The importance of the entire methodology based on the feasible adaptive weighted LADE is illustrated by simulation results and real data analysis on three U.S. economic data sets.