I have coded a model but it comes with NaN in exogenous processes. I used initval block, setting Xiss a very small value, so that any equations divided by it would not be NaN. But how am I supposed to do with the exogenous process? I have tried out shutting down all varexos, by setting them into zero at initval. But it fails.

Thanks for your quick reply. Now the problems turn to other equations, and I think it should be fixed through parameter calibration. Well, I have a quick question for calibration. In my reference paper that I use to code this model, I have got some estimated parameters for calibration. However, some parameters with factors of 400, such as steady state inflation, growth rate and others. Does that means that I have to use quarter value to calibrate the parameters? To be specific, if the steady state inflation is 2% annualized, then the calibrated parameter in the model \pi should be 0.005. Besides, other parameters like steady state interest rate should follow the suite. Right?

Yes, the model is calibrated in quarterly frequency, but the measured interest and inflation rates are annual. That’s why you have to match the quarterly steady state inflation rate and the discount factor to quarterly values and need an observation equation to map to the data.