The local asymptotic normality property is established fur a regression mod
el with fractional ARIMA(p, d, q) errors. This result allows for solving, i
n an asymptotically optimal way, a variety of inference problems In the lon
g-memory context: hypothesis testing, discriminant analysis, rank-based tes
ting, locally asymptotically minimax and adaptive estimation, etc. The prob
lem of testing linear constraints on the parameters, the discriminant analy
sis problem, and the construction of locally asymptotically minimax adaptiv
e estimators are treated in some detail.