SPCE Module - Discrepancy between uq_print output and model output moments


I have been using the new Stochastic PCE module to create surrogate models of a stochastic simulator and I recently noticed a discrepancy between the standard deviation output by calling the uq_print function and the standard deviation saved in the model structure. The following screenshot indicates the discrepancy as observed when working with my model.
As you can see, the discrepancy is quite significant!

I tested this on the SPCE example (Stochastic SIR) provided as well and noticed the same discrepancy. (I am unable to provide a screenshot of this due to the age of my account)

This seems quite odd, can you clarify if this is a mistake in my understanding or in the processing of the metamodel?

P.S. I did some more testing and found that if I compute as standard deviation as
mySPCE.SPCE.Moments.Var - mySPCE.SPCE.Sigma^2
I get the same result as is displayed in uq_print.

Dear @AGerkis

uq_print only reports the standard deviation of the SPCE (computed from the coefficients) whereas mySPCE.SPCE.Moments.Var stores the variance of the SPCE plus the noise level as introduced in the manual in Eq. 1.6:

Best regards

Hi Styfen,

Thanks for your response! So my understanding is that the value reported in the .Moments.Variance field is the “correct” value for interpreting the SPCE results?