maxLik

News

• An "author-created version" of the paper "maxLik: A package for maximum likelihood estimation in R" (Computational Statistics 26(3), 2011, p. 443-458, written by Arne Henningsen and Ott Toomet) is freely available at Arne's website: https://absalon.itslearning.com/data/ku/103018/publications/maxLik.pdf.
• Further News about the maxLik project are available at the project's R-Forge site.

What is maxLik?

• maxLik is an extension package for the "language and environment for statistical computing and graphics" called R.
• maxLik provides tools for maximum likelihood (ML) estimations.

What can maxLik do?

• (Likelihood) maximization using the following algorithms:
  • Newton-Raphson (NR). This method is based on quadratic approximation and uses both gradient and Hessian of the function.
  • Berndt-Hall-Hall-Hausman (BHHH). A version of NR where the Hessian is approximated by information equality (only works for maxmizing log-likelihood).
  • Broyden-Fletcher-Goldfarb-Shanno (BFGS). Another quasi-Newton method with a different approximation of Hessian.
  • conjugate gradients (CG). A method which only uses gradients (and function values) and does not approximate the Hessian. May be useful (but slow) for large problems.
  • simulated annealing (SANN). This method only uses function values. It is a stochastic optimization method which may be used if the objective function has secondary maximums, and in case of non-differentiable functions.
  • Nelder-Mead (NM). A simplex-based method using only the function values.
• Constrained maximization:
  • Linear equality and inequality constraints

Who has written maxLik?

• Ott Toomet
• Arne Henningsen

Where can I get maxLik?

• The released version is available on CRAN
  • You can download the source code, binaries for MacOS X and MS-Windows, and the reference manual from http://cran.r-project.org/web/packages/maxLik/index.html.
  • You can install the package by typing at the R prompt (if you have internet access):

     install.packages( "maxLik" )

• The current development version is available on R-Forge
  • You can download the source code and binaries for MacOS X and MS-Windows from http://r-forge.r-project.org/R/?group_id=254.
  • You can install the package by typing at the R prompt (if you have internet access):

     install.packages( "maxLik", repos="http://R-Forge.R-project.org" )

Under which license is maxLik released?

• Under the GNU General Public License (GPL)

Where can I ask questions, report bugs, or suggest new features?

• Please use a forum or "tracker" at the project's R-Forge site.