Releases: ikosmidis/MEstimation.jl
Releases · ikosmidis/MEstimation.jl
v0.2.0
MEstimation v0.1.0
MEstimation 0.1.0
The below are changes from GEEBRA v0.1.0 codebase, on which MEstimation was based on.
New functionality
- New
concentrate
keyword argument infit
forestimating_function_template
s, which allows adding bias-reducing adjustments only to a subset of estimating functions. - New
lower
andupper
keyword arguments infit
forobjective_function_template
s, which allows estimation in constrained parameters spaces. - New
regularizer
keyword argument infit
to allows for user-supplier regularizer functions. - New
slice
method for computing one-dimensional slices of objective and estimating functions. - Keyword arguments can be passed directly to
Optim.optimize
(e.g.autodiff = :forward
) through thefit
interface forobjective_function_template
s.
Bug fixes
objective_function
andestimating_function
are fully differentiable.
Other improvements, updates and additions
- The default output from
fit
now reports whether the optimization algorithm converged or not, and details on the (regularized) objective or estimating equations that are used. - Documentation written from scratch, and updated example in online documentation.
- New tests.
- Run time optimizations, mainly through using
DiffResults
, codebase refactoring, and explicit type specification. - Updates in compatibility with dependencies.