Skip to content
Snippets Groups Projects

smoother2histval: trap cases where both point and posterior estimates are available

Closed Marco Ratto requested to merge rattoma:bits-and-pieces into master

Merge request reports

Loading
Loading

Activity

Filter activity
  • Approvals
  • Assignees & reviewers
  • Comments (from bots)
  • Comments (from users)
  • Commits & branches
  • Edits
  • Labels
  • Lock status
  • Mentions
  • Merge request status
  • Tracking
  • Author Developer

    I added another commit, this time related to user defined mode_file option which is not preserved in dynare_estimation [re-set to empty].

  • @rattoma I have never used this feature, but when working on #1395 I was wondering how this affects loading smoother results and histval-files? I particularly envision cases where someone switches the loglinear-option on or off. In this case, we need to know whether the saved/stored values are in logs or not. #1395 takes the position that histval should always store the level, not the logged value and that the log is then taken whenever loglinear is specified. Do you know what this implies for the current ticket?

  • Author Developer

    @JohannesPfeifer smoother2histval uses SmoothedVariables and SmoothedShocks. As soon as the loglinear option is properly treated for SmoothedVariables, i.e. SmoothedVarables stores always levels, I think this should imply that the histval generated will always store the levels. However, looking at store_smoother_results, it seems that SmoothedVariables would instead include logged varibles. Which means histval will provide logged as well... Following your reasoning, we should then explicitly allow for a loglinear option in moother2histval, where histval is the exponent of smoothed variables if needed?

  • Exactly. We would need to store an indicator whether loglinear was used or not and then adjust smoother2histval accordingly.

  • Author Developer

    @JohannesPfeifer another small bit added to this PR

  • Manually merged.

  • mentioned in issue #1407 (closed)

Please register or sign in to reply
Loading