Finding peaks in geochemical distributions: A re-examination of the helium-continental crust correlation

Parman has recently suggested that a correlation exists between peaks in the ocean island basalt (OIB) 4He/3He distribution and peaks in crustal zircon ages. This correlation is based on matching peaks seen in smooth kernel density estimates. Kernel density estimation is a very useful technique, but care is required when choosing the smoothing bandwidth as spurious peaks can be produced if the bandwidth is too small. Here I provide an introduction to a general statistical technique for determining whether peaks in density estimates are significant, know as SiZer, focusing on its application to the 4He/3He data. SiZer identifies only two statistically significant peaks in the OIB 4He/3He distribution, compared with the eight peaks identified by Parman. The helium-continental crust correlation does not seem to be supported by the current data.