Synthetic Likelihood (SL) is a statistical methodology, introduced by Wood (2010), which can be used to do inference for models where the likelihood function is unavailable or intractable. SL is general-purpose in the sense that, as long as the user is able to simulate data from his model, he should be able to fit it. For this reason, the synlik R package puts very few limitations on the model being fitted: the user has simply to provide a simulator and (optionally) a function that transforms that data into summary statistics.
The package has been written with attention to computational efficiency and most of the functions provided support computation on multiple cores. At the moment synlik includes a flexible framework and set of simple, but useful and well documented statistical tools. In the future the package will be often updated and improved by the addition of new statistical methods and tools.
To check what synlik does, download of install it from here or from CRAN and have a look at its vignette here or type:
on the R console.
- Matteo Fasiolo, Simon Wood. An introduction to synlik (2014). R package version 0.1.0.
- Simon N Wood. Statistical inference for noisy nonlinear ecological dynamic systems. Nature, 466(7310):1102--1104, 2010.
To report bugs or the suggest new features, please contact me at matteo_DOT_fasiolo_AT_gmail.com.