This paper proposes a uniform inference method for the autoregressive coefficient of a dynamic panel model. Our method overcomes the asymptotic discontinuity of bias-corrected maximum livelihood (ML) between stationary and unit root parameter space. The proposed self-normalized t-statistic can be used to construct a confidence set that provides a uniformly valid asymptotic coverage over the whole parameter space.
Python script for applying the proposed method/simulation in the paper is available at my github repo: monchewharry/PaperLocal2Unity