I'm using the Moving Variance block in Simulink and it seems to be giving unexplained results for vectorised (muxed) inputs. My Simulink model is given below. I generate a random signal with power =1 (which equates to a variance of 1). I simulate for 10,000 samples and use a long window in the moving average block of 1000 so I should get the expeced variance (within a very small tolerance). As seen below in the middle case where I multiply the signal by 2, the varianace should change to 4, but it does not, nor does the variance of the straight signal (but it's OK in the block above). It seems that the vectorising is mucking up the calculations, but the help file seems to indicate that this should be acceptable.