v0.1.38

MacroModelling v0.1.38

Diff since v0.1.37

Major internal rework focused on nonlinear estimation. Codes are more flexible, faster, and more thoroughly tested.

Solution:

• Fast and robust QME, Sylvester, Lyapunov solutions
• Solution cache (starting point) for NSSS, QME, and Sylvester equations
• Expose solution algorithms for Lyapunov, Sylvester, and QME equations
• Improved NSSS solver: added Newton solver for speed
• Automatic homotopy continuation once NSSS solution was found
• Speedups thanks to custom sparse kronecker products
• Higher order solutions, stochastic steady states and log likelihood are reverse and forward mode AD compatible

Plotting:

• Improved model estimate plots
• More plotting options and possibility to add any plot attributes

Estimation:

• Added Cauchy prior
• Added nonlinear shock decomposition
• Corrected inversion filter for higher order solutions

Other:

• Completely overhauled testing: test all possible parameter combinations
• Improved model parsing
• verbose shows more internal calculations
• Redid internal derivatives
• Parameter derivatives are now always calculated and warning is thrown if they are expensive to calculate
• Added shock_size argument for IRFs
• dropped compatibility with Julia versions 1.8 and 1.9

Merged pull requests:

• Update plotting.jl (#80) (@LucaMingarelli)
• nonlinear estimation (#87) (@thorek1)

Closed issues:

• Macromodelling package issue (#95)
• Installation error (#96)