Show that the asymptotic relative efficiency of the sign or Wilcoxon signed-rank test relative to the t-test is unaltered by a linear change of variables - i.e., the ARE for a density $\displaystyle f(x)$ is the same as that for a density $\displaystyle g(x) = af(ax + b)$ where $\displaystyle a > 0$.

Not sure where to start here...