Risk simulation without calibration

[rkapl123]

I'm interested in finding out more about the possibilities of simulating scenarios that do not depend on risk-neutral market estimates (i.e. the default LGM model used by almost all examples). I've seen that there is an example utilizing the HWmodel instead of LGM, however there is little to no documentation on this in the userguide. Is there anything more definable than the sigma and kappa values? And is this also supposed to work with a multicurrency portfolio, in that case, how can the volatilities/reversions on the currencies be defined? On the other hand, could the LGM also be used without calibration and what would the input parameters look like there?

[xiangwang456]

Recently, while using sensitivity analysis or stress testing configuration scenarios, I found that when making an absolute adjustment of 1 basis point to the index curve, the result is not a 1 basis point adjustment to the original interest rate curve data. Instead, it is adjusting the data by 1 basis point on top of adjusting the original yield curve rates to continuously compounded rates. As a result, I had to adjust the continuously compounded interest rate data back to the original interest rate data based on the original interest rate's payment frequency before the applyshift() function, and then convert it back to continuously compounded data after applyshift() function. I'm unsure if there is a better method to handle this.

[rkapl123]

After the addition of historical/external simulation in 1.8.12, this is obsolete...