All Times ET
Keywords: Bias Correction, Bifurcating Autoregressive, Bootstrap, Confidence Intervals
The least-squares (LS) estimate of the autoregressive coefficient in the first order bifurcating autoregressive [BAR(1)] model is known to be biased, especially when the autoregressive parameter is close to the stationarity boundaries. In this study, we demonstrate the impact of bias correction on the behavior of bootstrap confidence intervals (CIs) for the BAR(1) autoregressive coefficient. Specifically, we examine the performance of asymptotic, percentile, and grid bootstrap CIs while correcting the LS bias by the linear bias function approach.