On Koul's minimum distance estimators in the regression models with long memory moving averages. (Q2574570)

The author considers a LRM where the error term is formed by a centered moving average stationary process with infinite memory, so called long-range memory case, where the sum of lagged correlations in the MA process is infinite. The paper objective consists in the estimation of regression parameters and proving their consistency and asymptotic normality. The estimation uses so called Koul's estimator which is of minimal distance estimator type, i.e., based on the minimization of a distance between the data and their prediction from the model (here it actually is a weighted regression median type estimator). It is necessary to say that the content of the paper is rather marginal, the definitions and result are concentrated on two pages, the rest are the proofs. Although they are not easy, one can follow the proofs of similar results, because the paper offers actually (a slight) extension of those of e.g.\ \textit{H.\ L.\ Koul} or \textit{H.\ L.\ Koul} and \textit{K.\ Mukherjee}. Nothing is said on the computational aspects of such an estimator, no practical (nor simulated) example is presented. It is also a pity that the work does not include also an analysis of the MA error term.