Skip to content
This repository has been archived by the owner. It is now read-only.

Trend prefiltering fixes #852

Conversation

Projects
None yet
2 participants
@JohannesPfeifer
Copy link
Contributor

commented Mar 1, 2015

Mostly fixes #679 and closes #255, #803, and #804.
Still to be done:

  • document new options once implementation is OKed
  • check treatment of various options and trends in smoother used in imcforecast.m
  • Check treatment of these issues in identification
  • Adjust convert_dyn_45_to_44 to reflect these changes
@JohannesPfeifer

This comment has been minimized.

Copy link
Contributor Author

commented Mar 4, 2015

Also fixes #780

@stepan-a

This comment has been minimized.

Copy link
Member

commented Mar 5, 2015

@JohannesPfeifer Thanks for splitting your changes into multiple commits. I still have a doubt about c32b1ec (#780). Whenever possible we should avoid to version binary files (because it makes no sense to track history of a binary file and because we need to know where the data are coming from to understand the test). In 2d3a80c could you replace the *.mat files by matlab scripts (or mod files) generating these files?

@JohannesPfeifer

This comment has been minimized.

Copy link
Contributor Author

commented Mar 5, 2015

@stepan-a I was also reluctant to add the binary mode-files. But it turned out that doing mode-finding with mode_compute=9 or mode_compute=10 took a long time given the size of the data. That is why I opted to include the saved mode-files. Given that there are 16 mod-files that require mode-finding, this would add about 8 hours to the run-time of the test-suite. Thus, I thought it would be preferable to provide the mode (they were generated by the same mod-files).

@stepan-a

This comment has been minimized.

Copy link
Member

commented Mar 5, 2015

@JohannesPfeifer Why would we need to use mode_compute=9, 10 for testing that there is no problem with profiteering of the data?

@JohannesPfeifer

This comment has been minimized.

Copy link
Contributor Author

commented Mar 5, 2015

I basically test whether the estimation runs away from the correct previous values. But you are correct that it makes more sense to start at the calibrated true value and run mode_compute=4. I changed it that way

@JohannesPfeifer JohannesPfeifer force-pushed the JohannesPfeifer:trend_prefiltering_fixes branch 7 times, most recently from c728486 to 5f3b83e May 20, 2015

@JohannesPfeifer JohannesPfeifer force-pushed the JohannesPfeifer:trend_prefiltering_fixes branch from 88465dd to 2127ba0 Jun 8, 2015

@JohannesPfeifer JohannesPfeifer force-pushed the JohannesPfeifer:trend_prefiltering_fixes branch from 2127ba0 to b700ae3 Jul 20, 2015

@JohannesPfeifer

This comment has been minimized.

Copy link
Contributor Author

commented Jul 20, 2015

@MichelJuillard @stepan-a From my perspective, this is the last big bugfix remaining for 4.5, but I guess there are still many kinks to be smoothed out. I now rebased it to reflect all the pull requests committed today. Any feedback is highly appreciated

@JohannesPfeifer JohannesPfeifer force-pushed the JohannesPfeifer:trend_prefiltering_fixes branch 2 times, most recently from cdb17bd to 25ceacd Jul 22, 2015

Johannes Pfeifer and others added some commits Mar 1, 2015

Factorize saving of smoother results
Also takes care of prefilter and loglinear option for calib_smoother
Closes #803 and #804
Add unit test for calibrated smoother
Tests loglinear, prefilter, and first_obs options
Remove binary files from unit tests
Start at true calibrated values instead (sufficient to see whether estimates run away)
Do not condition handing over of observable index on requesting forec…
…asts

Already the smoother uses them due to potential trends being present
Add constant to filtered and updated ML variables
Harmonizes treatment across ML and Bayesian estimation
Correct header of DsgeSmoother.m and write_smoother_results.m
Definition of updated variables was incorrect

@JohannesPfeifer JohannesPfeifer force-pushed the JohannesPfeifer:trend_prefiltering_fixes branch from 844451a to f953014 Aug 13, 2015

@stepan-a stepan-a force-pushed the DynareTeam:master branch from d60cc0a to 472f6c4 Mar 3, 2016

@stepan-a

This comment has been minimized.

Copy link
Member

commented Mar 23, 2016

Done in #1145

@stepan-a stepan-a closed this Mar 23, 2016

@JohannesPfeifer JohannesPfeifer deleted the JohannesPfeifer:trend_prefiltering_fixes branch Mar 23, 2016

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
You can’t perform that action at this time.