To protect people from electromagnetic field, Basic Restrictions (BR) are defined [1]. These BR fix a limit to be not exceeded. The metric associated with these BR is the Specific Absorption Rate (SAR). Reference Levels (RL) are also defined since the BR are difficult to check in situ. These RL set the maximum allowed electromagnetic field. The compliance to RL guaranties the compliance to BR. To evaluate the SAR in the human body, some anatomical models (phantoms) and numerical methods are used (e.g. Finite Difference in Time Domain). Based on this, studies show that for some configurations the Whole Body SAR (WBSAR) is close to BR. Other studies stressed the variability of the WBSAR due to the variability of human morphology [2]. Despite the computing resources development, the number of the phantoms is very limited. This limited number of phantoms does not allow using usual method such as Monte Carlo to assess the maximal threshold of the WBSAR for a given population. Hence the construction of a model of the WBSAR as a function of morphology is required. Nevertheless, the WBSAR is impacted by the external morphology (height and weight) and the internal morphology (proportion of fat, proportion of muscles...). But there is no statistical data concerning the internal ones. In this paper, the external morphology is focused and the internal morphology is released by considering homogeneous phantoms. A Bayesian sequential experiment planning is proposed. This method consists in refining the region of interest of the WBSAR statistical distribution for a given population. This region of interest is the threshold of the WBSAR at 95% (WBSAR95). This study is conducted in the case of a plane wave vertically polarized and frontally oriented on phantoms. The incident power is equal to 1W/m². The frequency is fixed at 2.1GHz.