Good afternoon, we are trying to estimate the latent value of the VIX with the nonlinear Kalman filter (NLKF) and time-varying volatility. However, after running the NLKF, the oo_ directory does not give us the information we need, namely the following:
Please, what do you think we should do? We are attaching the codes and the data.
V.mod (1.5 KB)
Run_V.m (389 Bytes)
IN.xlsx (10.7 KB)
Thank you for your prompt reply. Please, is there currently any particle filter that will give us smoothing results?
Implementing posterior IRFs would not be very difficult, we just need to add posterior uncertainty looping around the IRF routine with higher order approximations (for vectors of parameters drawn in the estimated posterior distribution). We could do the same for other moments.
Adding a nonlinear filter is a much bigger task. We would need to find a way to efficiently store the distribution of the latent variables in each period of the sample (for filtering we only need to keep track of the distribution). We currently do not have plans to implement a nonlinear smoother in Dynare.
Thank you very much for your response.