1. 17 Dec, 2020 1 commit
  2. 25 Jun, 2020 1 commit
  3. 23 Jun, 2020 1 commit
    • Johannes Pfeifer's avatar
      mr_hessian.m · 3e725e32
      Johannes Pfeifer authored
      Do not save files by default; prevents horse-race write access crashes for fast mod-files under Windows
      3e725e32
  4. 20 Jan, 2020 1 commit
  5. 20 Dec, 2019 1 commit
    • Sébastien Villemot's avatar
      New local_state_space_iteration_k MEX, for nonlinear filters at k-order · 996bdd6c
      Sébastien Villemot authored and Stéphane Adjemian's avatar Stéphane Adjemian committed
      It applies the approximated policy function to a set of particles, using
      Dynare++ routines.
      
      There is support for parallelization, using Dynare++ multithreading
      model (itself based on C++11 threads; we don’t use OpenMP because it is
      incompatible with MKL). For the time being, default to a single thread. This
      should be later refined through empirical testing.
      996bdd6c
  6. 19 Dec, 2019 1 commit
    • Marco Ratto's avatar
      provisions for making shock decompositions for epilogue variables. · 31c29d08
      Marco Ratto authored
      In case the epilogue formula is non-linear, the non additive non-linear term is distributed proportionally to the size of the individual shock contribution.
      It is triggered by new option with_epilogue, applicable to commands:
      1) shock_decomposition, realtime_shock_decomposition,
      where preprocessor should trigger
      options_.shock_decomp.with_epilogue=true;
      2) initial_condition_decomposition
      where preprocessor should trigger
      options_.initial_condition_decomp.with_epilogue=true;
      31c29d08
  7. 12 Dec, 2019 1 commit
  8. 29 Nov, 2019 1 commit
  9. 09 Jul, 2019 1 commit
  10. 27 Jun, 2019 2 commits
  11. 25 Jun, 2019 1 commit
  12. 18 Apr, 2019 2 commits
  13. 02 Apr, 2019 1 commit
  14. 19 Mar, 2019 1 commit
  15. 25 Oct, 2018 1 commit
  16. 19 Jul, 2018 1 commit
  17. 27 Jun, 2018 1 commit
  18. 28 May, 2018 2 commits
  19. 16 May, 2018 2 commits
    • Sébastien Villemot's avatar
      Improve fallback code when mjdgges DLL is absent · 59ad26a3
      Sébastien Villemot authored
      The new code relies on qz(..., 'real'), ordqz and ordeig, and returns a real
      decomposition. The previous version was using Sims' qzdiv and returned a
      complex decomposition.
      
      As a consequence, we can drop options_.qzdiv, which was used to detect when
      imaginary parts had to be dropped.
      
      This code does not work on Octave for the time being, but this is acceptable
      since it is only a fallback.
      59ad26a3
    • Stéphane Adjemian's avatar
      Added new option mh_tune_jscale. · f53be721
      Stéphane Adjemian authored
      Works only with the Random Walk Metropolis Hastings algorithm.
      
      Closes #1598
      f53be721
  20. 09 Jan, 2018 1 commit
  21. 06 Oct, 2017 1 commit
  22. 05 Oct, 2017 1 commit
  23. 16 Aug, 2017 1 commit
  24. 23 May, 2017 1 commit
  25. 16 May, 2017 1 commit
  26. 26 Apr, 2017 1 commit
  27. 24 Mar, 2017 2 commits
  28. 17 Mar, 2017 6 commits
  29. 08 Feb, 2017 2 commits
    • Stéphane Adjemian's avatar
      Removed penalty_hessian routine. · 9fbef0c1
      Stéphane Adjemian authored
       + Code factorization.
       + Added an option for using the penalized objective when computing numerically
       the hessian at the mode.
      
      Previous behaviour (introduced with penalty_hessian routine) was to compute the
      hessian matrix at the mode with the penalized objective function (instead of
      the original objective function). This behaviour hides problematic situations,
      where the computed hessian (using the original objective) would not be full
      rank. For instance, if the estimation ends up with a parameter on (or very
      close to) the bounds of its possible values (which is often not a desirable
      outcome), the estimated posterior variance would be zero for this
      parameter (with the original objective) because the hessian is not finite in
      this direction, while the posterior variance would be positive if the penalized
      objective is used instead. But this estimate would not be reliable by
      construction of the penalty which is quite ad-hoc (more fundamentally I do not
      think that there exists any rational for approximating the covariance matrix
      with the inverse of the hessian matrix if the mode is on the border of the set
      of possible values).
      
      This commit restore the behaviour previous to 2446ab02.
      
      An option is available for computing the hessian with the penalized
      objective function.
      9fbef0c1
    • Stéphane Adjemian's avatar
      Added new optimization routine (particleswarm). · 5b73d7d5
      Stéphane Adjemian authored
      Only available under Matlab if the Global Optimization Toolbox is installed.
      5b73d7d5