We discuss how prior regression on seasonal dummies leads to singulari
ties in log-periodogram regression procedures for the detection of lon
g memory. We suggest a modified procedure. We illustrate the problems
using monthly inflation data from [Hassler and Welters, 1995, Long Mem
ory in Inflation Rates: International Evidence, Journal of Business an
d Economic Statistics 13, 37-46]. Our new estimates reconfirm their ba
sic finding. (C) 1997 Elsevier Science S.A.