GitLab

Johannes Pfeifer authored Nov 15, 2018 and Stéphane Adjemian committed Dec 15, 2018

Also documents the behavior

Stéphane Adjemian authored Dec 15, 2018 and Stéphane Adjemian committed Dec 15, 2018

Only tests first order moemnt, second order moment and Cumulative Distribution
Function (should also add some tests for independance).

Stéphane Adjemian authored Dec 15, 2018 and Stéphane Adjemian committed Dec 15, 2018

Give access to all the implemented algorithms through the third argument. The
last argument is a structure with two fields `large` and `small`. The first
field specifies the algorithm to be used for α>1 while the second field defines
the algorithm to be used for α ∈ (0,1).

Default algorithms are:

 - Cheng (1977) for α>1,
 - Johnk (1964)  α ∈ (0,1).

Stéphane Adjemian authored Dec 14, 2018 and Stéphane Adjemian committed Dec 15, 2018

The description given by Devroye (1986) is clearly wrong, this implementation
is based on the original paper Väduva (Series Statistics, 1977).

Stéphane Adjemian authored Dec 14, 2018 and Stéphane Adjemian committed Dec 15, 2018

 - Definition of Y was wrong (error in Devroye's book).
 - Definition of X was wrong.

Also simplified the code w.r.t. update of the vector of indices.

Stéphane Adjemian authored Dec 12, 2018 and Stéphane Adjemian committed Dec 15, 2018

Also fixed implementation of Jonk's algorithm.

The binary is now picked from the Debian package.

We rather rely on the Gitlab setting that prevents some runners from accepting
jobs without tags.

In the *_params_derivs.m files, in the "hp" matrix, the preprocessor now fully
outputs the symmetric elements, so getH needs to be adapted.

By the way, also clean up code for "rpp" and "gpp", were full symmetric
elements were also already present (and therefore do not need to be duplicated
in getH).

kim2.mod: trigger case where estim_params_block is not detected

See merge request Dynare/dynare!1638

Ref Dynare/preprocessor#13

Ref Dynare/preprocessor#13

identification_analysis.m: Fix bug introduced in 60d95b65

See merge request Dynare/dynare!1637

We already used a variable called offset

evaluate_steady_state_file.m: Remove redundant setting of auxiliary variables

Closes #1175

See merge request Dynare/dynare!1635

The original problems seems to have been fixed by #1133

Essentially reverts Dynare/dynare@3c7e60b7 and closes #1175

evaluate_steady_state_file.m: fix two bugs in debugging mode

See merge request Dynare/dynare!1634

Johannes Pfeifer authored Nov 15, 2018 and Sébastien Villemot committed Nov 16, 2018

1. Rely on actually computed standard deviations in bayestopt_ instead of potentially unset estim_params_
2. Remove arbitrary normalizations/omitted normalization in case of division by 0 in normalization
3. Distinguish between 0 identification and division by 0 due to normalization in plots

(cherry picked from commit 7341e21a)