Estimating DSGE Model

I have two questions about estimating a dsge model.

  1. If I have express my data in logs, do i need to log linearize the model to estimate it or can i leave it in it’s non-linear state.

  2. I was looking at the replication attempt for Smets and Wouters, and wondering if the model is initially calibrated to solve for numerical steady state values which are used in initval. If yes, then would the calibrated parameters in matlab (I use matlab to recursively solve for steady state) affect the estimated parameters.

Thank you.

  1. Please refer to Pfeifer(2013): “A Guide to Specifying Observation Equations for the Estimation of DSGE Models”
  2. A proper replication does account for such parameter dependence via e.g. a steady_state_model-block. See