The aim of this contribution is to present a whole experimental methodology to assess the effects of uncertainties for a classical EMC issue. To this end, a cabinet was designed and achieved with controlled geometrical parameters including moving external slot and internal plate, an inner rotating unit embedded with stirrer and cable. From industrial expectations, the experimental setup and the theoretical basis will be described briefly. Then, the validation (convergence, accuracy and robustness) of the proposed stochastic methods will be obtained facing Monte Carlo (MC) measurements including three random parameters. Finally, the combination of stochastic techniques with sensitivity study will improve the global process. Physical and industrial contextsIn the ElectroMagnetic Compatibility (EMC) literature, designing electronic large systems is mostly based upon "worst" cases approaches. In accordance with standards, this mainly sets two problems: the need for precise and efficient tools to quantify more realistic EMC margins, jointly with trustworthy reliability levels. Non-exhaustive stateof-the-art EMC stochastic issues contain different philosophies to integrate this problem for instance involving Printed Circuit Boards (PCBs) [1], cable coupling [2-3] and effects of uncertain High Intensity Radiated Fields (HIRFs) [3]. Materials and methods Overview of stochastic measurements setupsThe study presented in [4] detailed the design and achievement of a box including different geometrical parameters acting as Random Variables (RV). The governing idea was to manufacture a device providing mechanically various and precise configurations of the equipment. This allows defining different EMC classical setups involving different devices and outputs. Indeed, as depicted respectively in Fig.2-A and Fig.3-A, slots and inner volume modifications are allowed, and a rotating unit enables modifications around the location of different equipment (cable for instance). First, a spectrum analyzer Anritsu MS 2663C (bandwidth 9kHz-3GHz) has been used for power measurement ( Figure 2). Then, S 12 parameter (Figures 1 and 2) between emission and reception antennas is measured by a network analyzer Rohde & Schwarz ZVB 8 (bandwidth 300kHz-8GHz). A common strategy has been defined since the measurements were achieved in one go (no break); this has required a whole automation of the process (especially for MC treatments). Due to the similarity of the MC and Stochastic Collocation (SC) approaches (details in section 2.2), the statistical treatments required only to use the chosen points from this database. Stochastic theory and sensitivity analysis needsAccording to the SC foundations [4], the technique is close to MC philosophy since it only requires a smart sampling of input data and to straightforward apply the deterministic setup (numerical and/or experimental) from this set of points. The theoretical details and some chosen examples of SC weighted points sets may be found in [5] where the limitations of the method were presented. Du...