Convex combinations of long memory estimates from different sampling rates (Q2463650)

A procedure for obtaining a convex combination of long-memory estimates from the same data observed at different sampling rates is implemented. Convexly combining the estimates decreases the estimation variance while introducing a bias. The authors apply corrections for this bias in the memory parameter to the estimates from decreasing sampling rates using the heuristic bias formula of \textit{L. R. Souza} and \textit{J. Smith} [Bias in the memory parameter for different sampling rates. Int. J. Forecast 18, 299--313 (2002)]. It is shown that the root mean square error of the estimation is reduced on the average by about 30 \% for \(ARFIMA (0,d,0)\) processes.