estimate: Model Estimation

In tsmodels/tsets: Exponential Smoothing Models

Model Estimation

Description

Estimates a model given a specification object using maximum likelihood.

Usage

## S3 method for class 'tsets.spec'
estimate(
  object,
  solver = "nlminb",
  control = list(trace = 0),
  autodiff = TRUE,
  ...
)

Arguments

object	an object of class “tsets.spec”.
solver	one of either “solnp”, “nlminb” or “optim”. The latter uses the L-BFGS-B algorithm from the lbfgsb3c package. For option “autodiff”, valid solvers are “nlminb” and “nloptr”.
control	solver control parameters.
autodiff	whether to use automatic differentiation for estimation. This makes use of the tsetsad package.
...	only additional argument which can be passed when choosing autodiff is that of “use_hessian” which tells the solver to make use of second derivatives.

Details

The maximum likelihood estimation uses bound constraints for some of the parameters as described in the online book of tsmodels. Additionally, for parameters which are constrained to be less than another parameter (e.g. beta<alpha), a simple barrier approach is adopted which adjusts the previous likelihood value upwards by some fixed percentage of that value during the minimization. The observation variance is not directly estimated but instead concentrated out of the likelihood. When autodiff is TRUE with the nloptr solver, the constraints and their jacobian are explicitly used, whilst a soft barrier constraint is used in the case of the nlminb solver.

Value

An object of class “tsets.estimate”

tsmodels/tsets documentation built on Oct. 8, 2022, 9:15 a.m.