International guidelines/standards for human exposure to electromagnetic fields have recently been revised to update the dosimetric reference limits (DRLs or basic restrictions) and exposure reference levels (ERLs), specifically for frequencies above 6 GHz. At such frequencies, the ERL is defined in terms of incident power density (IPD) and used as a practical quantity to assess compliance with DRLs (absorbed or epithelial power density) and therefore appropriately limits temperature elevation at the body surface. In the exposure standards, IPD is spatially averaged over an area of 4 cm<sup>2</sup> below 30 GHz and 1 cm<sup>2</sup> above 30 GHz, however the definition of IPD is given in a theoretical manner. With the progress in the development of product safety compliance assessment standards, one concern has been how to define the IPD considering practical measurement procedures. Two definitions or averaging methods were considered: using IPD vectors normal to the averaging surface and using magnitude (norm) of IPD vectors. As the exposure guidelines are intended to prevent excessive tissue heating, statistical analysis was therefore undertaken to investigate which IPD metric better correlates with the temperature increase. To this end, a large data set for several exposure scenarios was collected by different research institutions. The analysis of the obtained results is presented and shows that both definitions have high correlation with temperature rise, with slightly better correlation (0.9 vs. 0.8) for the definition using the magnitude of IPD vectors.