mlysy.r-universe.dev

[mlysy] SuperGauss 2.0.4

mlysy@uwaterloo.ca (Martin Lysy) — Tue, 09 Sep 2025 17:26:16 GMT

Likelihood evaluations for stationary Gaussian time series are typically obtained via the Durbin-Levinson algorithm, which scales as O(n^2) in the number of time series observations. This package provides a "superfast" O(n log^2 n) algorithm written in C++, crossing over with Durbin-Levinson around n = 300. Efficient implementations of the score and Hessian functions are also provided, leading to superfast versions of inference algorithms such as Newton-Raphson and Hamiltonian Monte Carlo. The C++ code provides a Toeplitz matrix class packaged as a header-only library, to simplify low-level usage in other packages and outside of R.

[mlysy] LocalCop 0.0.2

mlysy@uwaterloo.ca (Martin Lysy) — Tue, 24 Sep 2024 19:46:15 GMT

Implements a local likelihood estimator for the dependence parameter in bivariate conditional copula models. Copula family and local likelihood bandwidth parameters are selected by leave-one-out cross-validation. The models are implemented in 'TMB', meaning that the local score function is efficiently calculated via automated differentiation (AD), such that quasi-Newton algorithms may be used for parameter estimation.

[mlysy] mniw 1.0.2

mlysy@uwaterloo.ca (Martin Lysy) — Thu, 19 Sep 2024 16:05:44 GMT

Density evaluation and random number generation for the Matrix-Normal Inverse-Wishart (MNIW) distribution, as well as the the Matrix-Normal, Matrix-T, Wishart, and Inverse-Wishart distributions. Core calculations are implemented in a portable (header-only) C++ library, with matrix manipulations using the 'Eigen' library for linear algebra. Also provided is a Gibbs sampler for Bayesian inference on a random-effects model with multivariate normal observations.

[mlysy] optimCheck 1.0.1

mlysy@uwaterloo.ca (Martin Lysy) — Thu, 05 Sep 2024 19:20:53 GMT

Tools for checking that the output of an optimization algorithm is indeed at a local mode of the objective function. This is accomplished graphically by calculating all one-dimensional "projection plots" of the objective function, i.e., varying each input variable one at a time with all other elements of the potential solution being fixed. The numerical values in these plots can be readily extracted for the purpose of automated and systematic unit-testing of optimization routines.

[mlysy] MADPop 1.1.6

mlysy@uwaterloo.ca (Martin Lysy) — Mon, 26 Feb 2024 17:23:21 GMT

Tools for the analysis of population differences using the Major Histocompatibility Complex (MHC) genotypes of samples having a variable number of alleles (1-4) recorded for each individual. A hierarchical Dirichlet-Multinomial model on the genotype counts is used to pool small samples from multiple populations for pairwise tests of equality. Bayesian inference is implemented via the 'rstan' package. Bootstrapped and posterior p-values are provided for chi-squared and likelihood ratio tests of equal genotype probabilities.

[mlysy] nicheROVER 1.1.2

mlysy@uwaterloo.ca (Martin Lysy) — Fri, 13 Oct 2023 15:50:55 GMT

Implementation of a probabilistic method to calculate 'nicheROVER' (_niche_ _r_egion and niche _over_lap) metrics using multidimensional niche indicator data (e.g., stable isotopes, environmental variables, etc.). The niche region is defined as the joint probability density function of the multidimensional niche indicators at a user-defined probability alpha (e.g., 95%). Uncertainty is accounted for in a Bayesian framework, and the method can be extended to three or more indicator dimensions. It provides directional estimates of niche overlap, accounts for species-specific distributions in multivariate niche space, and produces unique and consistent bivariate projections of the multivariate niche region. The article by Swanson et al. (2015) provides a detailed description of the methodology. See the package vignette for a worked example using fish stable isotope data.

[mlysy] msde 1.0.5

mlysy@uwaterloo.ca (Martin Lysy) — Tue, 24 May 2022 15:39:05 GMT

Implements an MCMC sampler for the posterior distribution of arbitrary time-homogeneous multivariate stochastic differential equation (SDE) models with possibly latent components. The package provides a simple entry point to integrate user-defined models directly with the sampler's C++ code, and parallelizes large portions of the calculations when compiled with 'OpenMP'.

[mlysy] LMN 1.1.2

mlysy@uwaterloo.ca (Martin Lysy) — Fri, 25 Feb 2022 01:35:11 GMT

Efficient Frequentist profiling and Bayesian marginalization of parameters for which the conditional likelihood is that of a multivariate linear regression model. Arbitrary inter-observation error correlations are supported, with optimized calculations provided for independent-heteroskedastic and stationary dependence structures.

[mlysy] rdoxygen 2.0.0

clemens@nevrome.de (Clemens Schmid) — Wed, 04 Sep 2019 13:39:14 GMT

Create 'doxygen' documentation for source code in R packages, and optionally make it accessible as an R vignette. Includes a 'RStudio' Addin to easily trigger the doxygenize process.