“…In practice, even if the exact distribution of the innovations is unknown, as long as the data have similar tail behavior as the density function used in the estimation, inferences based on these methods still have good sampling performance. Thus, we may consider adaptive (Hansen and Lee, 1994;Seo, 1996;Beelders, 1998) or partially adaptive (Bickel, 1982, p. 664;Potscher and Prucha, 1986;Xiao, 1999) estimation methods so that the data density can be approximated. For example, to capture the feature of heavy tails in economic and financial data, we may consider a partially adaptive estimator based on the Student-t distributions, which has wide applications in economic analysis.…”