This package provides tools for maximum likelihood estimation of parameters of scientific models. Based on Goffe et al (1994) <doi:10.1016/0304-4076(94)90038-8>.

Select statistically similar research groups by backward selection using various robust algorithms, including a heuristic based on linear discriminant analysis, multiple heuristics based on the test statistic, and parallelized exhaustive search.

Fast implementations to compute the genetic covariance matrix, the Jaccard similarity matrix, the s-matrix (the weighted Jaccard similarity matrix), and the (classic or robust) genomic relationship matrix of a (dense or sparse) input matrix (see Hahn, Lutz, Hecker, Prokopenko, Cho, Silverman, Weiss, and Lange (2020) <doi:10.1002/gepi.22356>). Full support for sparse matrices from the R-package Matrix'. Additionally, an implementation of the power method (von Mises iteration) to compute the largest eigenvector of a matrix is included, a function to perform an automated full run of global and local correlations in population stratification data, a function to compute sliding windows, and a function to invert minor alleles and to select those variants/loci exceeding a minimal cutoff value. New functionality in locStra allows one to extract the k leading eigenvectors of the genetic covariance matrix, Jaccard similarity matrix, s-matrix, and genomic relationship matrix via fast PCA without actually computing the similarity matrices. The fast PCA to compute the k leading eigenvectors can now also be run directly from bed'+'bim'+'fam files.

Fast binning of multiple variables using parallel processing. A summary of all the variables binned is generated which provides the information value, entropy, an indicator of whether the variable follows a monotonic trend or not, etc. It supports rebinning of variables to force a monotonic trend as well as manual binning based on pre specified cuts. The cut points of the bins are based on conditional inference trees as implemented in the partykit package. The conditional inference framework is described by Hothorn T, Hornik K, Zeileis A (2006) <doi:10.1198/106186006X133933>.

Implementations of estimation algorithm of low rank plus sparse structured VAR model by using Fast Iterative Shrinkage-Thresholding Algorithm (FISTA). It relates to the algorithm in Sumanta, Li, and Michailidis (2019) <doi:10.1109/TSP.2018.2887401>.

This package provides modular, graph-based agents powered by large language models (LLMs) for intelligent task execution in R. Supports structured workflows for tasks such as forecasting, data visualization, feature engineering, data wrangling, data cleaning, SQL', code generation, weather reporting, and research-driven question answering. Each agent performs iterative reasoning: recommending steps, generating R code, executing, debugging, and explaining results. Includes built-in support for packages such as tidymodels', modeltime', plotly', ggplot2', and prophet'. Designed for analysts, developers, and teams building intelligent, reproducible AI workflows in R. Compatible with LLM providers such as OpenAI', Anthropic', Groq', and Ollama'. Inspired by the Python package langagent'.

Temporary and permanent message queues for R. Built on top of SQLite databases. SQLite provides locking, and makes it possible to detect crashed consumers. Crashed jobs can be automatically marked as "failed", or put in the queue again, potentially a limited number of times.

Computational routines for estimating local Gaussian parameters. Local Gaussian parameters are useful for characterizing and testing for non-linear dependence within bivariate data. See e.g. Tjostheim and Hufthammer, Local Gaussian correlation: A new measure of dependence, Journal of Econometrics, 2013, Volume 172 (1), pages 33-48 <DOI:10.1016/j.jeconom.2012.08.001>.

This package provides tools to decompose differences in cohort health expectancy (HE) by age and cause using longitudinal data. The package implements a novel longitudinal attribution method based on a semiparametric additive hazards model with time-dependent covariates, specifically designed to address interval censoring and semi-competing risks via a copula framework. The resulting age-cause-specific contributions to disability prevalence and death probability can be used to quantify and decompose differences in cohort HE between groups. The package supports stepwise replacement decomposition algorithms and is applicable to cohort-based health disparity research across diverse populations. Related methods include Sun et al. (2023) <doi:10.1177/09622802221133552>.

Estimate and confidence/credible intervals for an unknown regressor x0 given an observed y0.

The proposed method aims at predicting the longitudinal mean response trajectory by a kernel-based estimator. The kernel estimator is constructed by imposing weights based on subject-wise similarity on L2 metric space between predictor trajectories as well as time proximity. Users could also perform variable selections to derive functional predictors with predictive significance by the proposed multiplicative model with multivariate Gaussian kernels.

Combines Latent Dirichlet Allocation (LDA) and Bayesian multinomial time series methods in a two-stage analysis to quantify dynamics in high-dimensional temporal data. LDA decomposes multivariate data into lower-dimension latent groupings, whose relative proportions are modeled using generalized Bayesian time series models that include abrupt changepoints and smooth dynamics. The methods are described in Blei et al. (2003) <doi:10.1162/jmlr.2003.3.4-5.993>, Western and Kleykamp (2004) <doi:10.1093/pan/mph023>, Venables and Ripley (2002, ISBN-13:978-0387954578), and Christensen et al. (2018) <doi:10.1002/ecy.2373>.

This package provides functions for simulating missing morphometric data randomly, with taxonomic bias and with anatomical bias. LOST also includes functions for estimating linear and geometric morphometric data.

Collections of functions allowing random number generations and estimation of Liouville copulas, as described in Belzile and Neslehova (2017) <doi:10.1016/j.jmva.2017.05.008>.

Compute power and sample size for linear models of longitudinal data. Supported models include mixed-effects models and models fit by generalized least squares and generalized estimating equations. The package is described in Iddi and Donohue (2022) <DOI:10.32614/RJ-2022-022>. Relevant formulas are derived by Liu and Liang (1997) <DOI:10.2307/2533554>, Diggle et al (2002) <ISBN:9780199676750>, and Lu, Luo, and Chen (2008) <DOI:10.2202/1557-4679.1098>.

When the values of the outcome variable Y are either 0 or 1, the function lsm() calculates the estimation of the log likelihood in the saturated model. This model is characterized by Llinas (2006, ISSN:2389-8976) in section 2.3 through the assumptions 1 and 2. The function LogLik() works (almost perfectly) when the number of independent variables K is high, but for small K it calculates wrong values in some cases. For this reason, when Y is dichotomous and the data are grouped in J populations, it is recommended to use the function lsm() because it works very well for all K.

Create maps made of lines. The package contains one function: linemap(). linemap() displays a map made of lines using a raster or gridded data.

This package provides a single analysis path that includes distance-based ordination, global tests of any effect of the microbiome, and tests of the effects of individual taxa with false-discovery-rate (FDR) control. It accommodates both continuous and discrete covariates as well as interaction terms to be tested either singly or in combination, allows for adjustment of confounding covariates, and uses permutation-based p-values that can control for sample correlations. It can be applied to transformed data, and an omnibus test can combine results from analyses conducted on different transformation scales. It can also be used for testing presence-absence associations based on infinite number of rarefaction replicates, testing mediation effects of the microbiome, analyzing censored time-to-event outcomes, and for compositional analysis by fitting linear models to centered-log-ratio taxa count data.

Bayesian population size estimation using non parametric latent-class models.

Use of this package is deprecated. It has been renamed to LifeInsureR'.

Fits and tests logistic joinpoint models.

L-systems or Lindenmayer systems are parallel rewriting systems which can be used to simulate biological forms and certain kinds of fractals. Briefly, in an L-system a series of symbols in a string are replaced iteratively according to rules to give a more complex string. Eventually, the symbols are translated into turtle graphics for plotting. Wikipedia has a very good introduction: en.wikipedia.org/wiki/L-system This package provides basic functions for exploring L-systems.

This package contains a set of functions to create data libraries, generate data dictionaries, and simulate a data step. The libname() function will load a directory of data into a library in one line of code. The dictionary() function will generate data dictionaries for individual data frames or an entire library. And the datestep() function will perform row-by-row data processing.

Supervised learning techniques designed for the situation when the dimensionality exceeds the sample size have a tendency to overfit as the dimensionality of the data increases. To remedy this High dimensionality; low sample size (HDLSS) situation, we attempt to learn a lower-dimensional representation of the data before learning a classifier. That is, we project the data to a situation where the dimensionality is more manageable, and then are able to better apply standard classification or clustering techniques since we will have fewer dimensions to overfit. A number of previous works have focused on how to strategically reduce dimensionality in the unsupervised case, yet in the supervised HDLSS regime, few works have attempted to devise dimensionality reduction techniques that leverage the labels associated with the data. In this package and the associated manuscript Vogelstein et al. (2017) <arXiv:1709.01233>, we provide several methods for feature extraction, some utilizing labels and some not, along with easily extensible utilities to simplify cross-validative efforts to identify the best feature extraction method. Additionally, we include a series of adaptable benchmark simulations to serve as a standard for future investigative efforts into supervised HDLSS. Finally, we produce a comprehensive comparison of the included algorithms across a range of benchmark simulations and real data applications.