It is important to note that the scaling coefficient of the
equivalent Normal variates must be estimated through the
ranks of the data, as discussed above. Another important issue
is the treatment of apparent trends in the data. In the
case of autocorrelated data, the existence of trends will affect
the estimates of the autocorrelation in the data, which
adversely affects the power of the test. It is also true that
trends will affect the estimation of the scaling coefficient
H as shown by numerous studies. Due to this fact, Hamed
and Rao (1998) recommend the removal of apparent trends
using a suitable non-parametric trend estimator before the
estimation of the autocorrelation. A detailed investigation