Third-order derivatives with estimation command

[amg]

Hi, In the current version of Dynare, the estimation is done based on a first order approximation. Why would you need a third order approximation for estimation ? If you need the third order derivates you have to use the stoch_simul command instead.

Best, Stéphane.

[amg]

Hi,

E.g., to estimate time-varying risk premia, I'd need a third-order approximation.
I can provide Dynare with a suitable replacement for the linear estimation routines, but would hate to have to redo everything (i.e., if Dynare provides me with the third-order derivaitves within the estimation command that it is of course capable of producing, I can take care of everything else I need: I can retool DSGELikelihood.m etc. to my needs, but dynare_m.exe would be a little more involved). Starting with stoch_simul would require me to reprogram the entire estimation routines of Dynare as well as forcing--I believe--the rederivation of derivatives (i.e., _dynamic.m) at every parameter value iteration. Anyway, if you set order=2 in the estimation command, Dynare will provide the second order derivatives, setting order=3 still leaves you with only the second derivaitves of the policy function.

[StephaneAdjemian]

Hi,

I tried to use order=3 option in the estimation command for the fs2000 model, and I get the third order derivates (stored in the matrix v3)... I used dynare version 4.2.2.

Best,
Stéphane.

[amg]

Thanks, got it... user error on my part.