Density Estimation and Bump-Hunting by the Penalized Likelihood Method Exemplified by Scattering and Meteorite Data: Comment.

In their paper, Good and Gaskins determine empirically the number of terms required in Hermite and Fourier expansions to obtain equivalent degrees of approximation. The argument that a sign pattern in the coefficients in the Hermite expansion of a function slows the convergence is ingenious, but there is another reason for the better approximation by Fourier series than for Hermite series that is probably more important.