Mean square error

Hi all.

I have a confusion with the functionality of the dynare_sensitive command. I am trying to find the mean square error. I have found that the command dynare_sensitive works to find the rmse but I do not understand very well:

  1. To use this command, is it necessary to make Bayesian estimates of the parameters of the model?
  2. Does this command go before or after stoch_simul or where in the model does it go?
  3. Can I find the errors of the variables with another command?

I appreciate any contribution

  1. dynare_sensitivity computes rmse as one of the steps to carry out the sensitivity analysis as in Ratto’s computational economics paper I believe. Therefore, it doesn’t compute mse as output.
  2. Yes.
  3. It’s orthogonal I believe. You can run it after stoch_simul for sure.
  4. clarify your question a bit?

Maybe formulate what you want to find in math so other people can help you better

1 Like

The MSE is just an average squared deviation from some object. You can compute it for a variety of things. What is the exact object you are interested in?

I want to use it as an indicator to choose a policy rule


You still did not say, what

refers to.

I am evaluating different policy rules. Graph how each variable behaves before the shocks with each rule, it turns out that the graphs are very close visually, the difference is not seen. I needed to choose the best rule and I thought about using the mean square error as an indicator, something aspiring to measure the area under the curve. So I want to get the mean square error of each variable, I don’t know if this approach is wrong?
I hope to be clear

You again did not answer my question. Are you talking about the mean squared forecast error between actual data realizations and forecasts from the model?