We give expansions about the Gumbel distribution in inverse powers of
n and log
n for
M n , the maximum of a sample size
n or
n+1 when the
j-th observation is
\(\mu (\tfrac{j}{n}) + e_j \), µ is any smooth trend function and the residuals {
e j } are independent and identically distributed with
$$P(e > r) \approx \exp ( - \delta x)x^{d_0 } \sum\limits_{k = 1}^\infty {c_k x^{ - k\beta } } $$
as
x→∞. We illustrate practical value of the expansions using simulated data sets.