Clarification is needed regarding the calibration of stressed interest rates in the CEIOPS Calibration Paper (2010-04-15). PCA is applied to four different datasets, and a regression model is subsequently used on the derived (standardised) principal components (PCs), or factors, to calculate the "beta" sensitivity of each maturity. As described in the paper, the combined sum of these sensitivities is then used to derive the stress factor at the 99.5% level. I would like to understand which type of regression model was used, how the combined sum was calculated, and which assumptions (e.g. regarding the underlying distributions) were made in order to derive the stress factors at the 99.5% level. Furthermore, I would like to understand how this confidence level should be interpreted. Is it intended to represent the Value-at-Risk for each individual maturity? If so, the underlying distributional assumptions would be crucial for understanding the calibration methodology.