Question about correction/constant

[dsgeapprox]

Hi,
I have a question about the correction term in my second order approximation. I have a model with stochastic volatility that I am trying to approximate. The model has an error term that is being multiplied by an endogenous variable. The endogenous variable follows a log AR(1) process with a shock. When I use the approximation in Dynare, I find that changing the variance of the shock to the AR(1) process does not change the correction term. When I calculate the correction term on my own, I find that the correction term changes as the variance changes. Am I making a mistake in my coding in Dynare, or is the result I am getting correct in Dynare? (It's entirely possible that my calculation is correct.)

I have attached a mod file with a simplified version of my code. The "sig" variable is allowed to change over time when there are shocks to "u." When I change the variance of "u" (in line 27) I don't find that the correction term (and therefore the constant) for the approximation to "y" does not change like when I calculate this on my own. In this case, with var u=2, by my calculation the correction should be 1.4923 instead of 0.206612. I am wondering if I have a problem in my coding in the .mod file or if this is a type of model that isn't supported by Dynare or if this is indeed correct and my calculation is incorrect. Any help is appreciated.

Thanks in advance

[dsgeapprox]

Has anyone had a chance to look at this? Any input is appreciated. Thanks.

[jpfeifer]

Could you tell me how you computed the supposed outcome?