Shed light on black box machine learning models by the help of model performance, variable importance, global surrogate models, ICE profiles, partial dependence (Friedman J. H. (2001) <doi:10.1214/aos/1013203451>), accumulated local effects (Apley D. W. (2016) <doi:10.48550/arXiv.1612.08468>), further effects plots, interaction strength, and variable contribution breakdown (Gosiewska and Biecek (2019) <doi:10.48550/arXiv.1903.11420>). All tools are implemented to work with case weights and allow for stratified analysis. Furthermore, multiple flashlights can be combined and analyzed together.

This package provides a compact and explicit machine learning framework for supervised learning, resampling-based evaluation, hyperparameter tuning, learner comparison, interpretation, and plug-in g-computation. The package uses standard formulas for model specification and provides stable S3 interfaces for fitting, evaluation, tuning, interpretation, and causal estimation across a learner registry with multiple backend engines. Implemented interpretation methods build on established approaches such as permutation-based variable importance, partial dependence, individual conditional expectation, accumulated local effects, SHAP, and LIME; see Friedman (2001) <doi:10.1214/aos/1013203451>, Goldstein et al. (2015) <doi:10.1080/10618600.2014.907095>, Apley and Zhu (2020) <doi:10.1111/rssb.12377>, Lundberg and Lee (2017) <doi:10.48550/arXiv.1705.07874>, and Ribeiro et al. (2016) <doi:10.48550/arXiv.1602.04938>. The framework is intentionally opinionated: preprocessing is expected to occur outside the modeling step, and the API emphasizes explicit inputs, consistent object contracts, and compact interfaces rather than feature-by-feature competition with larger machine learning ecosystems.

This package provides a fast and scalable linear mixed-effects model (LMM) estimation algorithm for analysis of single-cell differential expression. The algorithm uses summary-level statistics and requires less computer memory to fit the LMM.

This package provides a wrapper for the python module FIORA as well as a shiny'-App to facilitate data processing and visualization. FIORA allows to predict Mass-Spectra based on the SMILES code of chemical compounds. It is described in the Nature Communications article by Nowatzky (2025) <doi:10.1038/s41467-025-57422-4>.

Non-Uniform Fast Fourier Transform ('NUFFT')-accelerated local polynomial regression and kernel density estimation for large, scattered, or complex-valued datasets. Provides automatic bandwidth selection via Generalized Cross-Validation (GCV) for regression and Likelihood Cross-Validation (LCV) for density estimation. This is the R port of the fastLPR MATLAB'/'Python toolbox, achieving O(N + M log M) computational complexity through custom NUFFT implementation with Gaussian gridding. Supports 1D/2D/3D data, complex-valued responses, heteroscedastic variance estimation, and confidence interval computation. Performance optimized with vectorized R code and compiled helpers via Rcpp'/'RcppArmadillo'. Extends the FKreg toolbox of Wang et al. (2022) <doi:10.48550/arXiv.2204.07716> with Python and R ports. Applied in Li et al. (2022) <doi:10.1016/j.neuroimage.2022.119190>. Uses NUFFT methods based on Greengard and Lee (2004) <doi:10.1137/S003614450343200X>, binning-accelerated kernel estimation of Wand (1994) <doi:10.1080/10618600.1994.10474656>, and local polynomial regression framework of Fan and Gijbels (1996, ISBN:978-0412983214).

Query data hosted in Microsoft Fabric'. Provides helpers to open DBI connections to SQL endpoints of Lakehouse and Data Warehouse items; submit Data Analysis Expressions ('DAX') queries to semantic model datasets in Microsoft Fabric and Power BI'; read Delta Lake tables stored in OneLake ('Azure Data Lake Storage Gen2'); and execute Spark code via the Livy API'.

This package implements methods for multiple change point detection in multivariate time series with non-stationary dynamics and cross-correlations. The methodology is based on a model in which each component has a fluctuating mean represented by a random walk with occasional abrupt shifts, combined with a stationary vector autoregressive structure to capture temporal and cross-sectional dependence. The framework is broadly applicable to correlated multivariate sequences in which large, sudden shifts occur in all or subsets of components and are the primary targets of interest, whereas small, smooth fluctuations are not. Although random walks are used as a modeling device, they provide a flexible approximation for a wide class of slowly varying or locally smooth dynamics, enabling robust performance beyond the strict random walk setting.

This package provides functional control charts for statistical process monitoring of functional data, using the methods of Capezza et al. (2020) <doi:10.1002/asmb.2507>, Centofanti et al. (2021) <doi:10.1080/00401706.2020.1753581>, Capezza et al. (2024) <doi:10.1080/00224065.2024.2383674>, Capezza et al. (2024) <doi:10.1080/00401706.2024.2327346>, Centofanti et al. (2025) <doi:10.1080/00224065.2024.2430978>, Capezza et al. (2025) <doi:10.48550/arXiv.2410.20138>. The package is thoroughly illustrated in the paper of Capezza et al (2023) <doi:10.1080/00224065.2023.2219012>.

Statistical tool set for population genetics. The package provides following functions: 1) empirical Bayes estimator of Fst and other measures of genetic differentiation, 2) regression analysis of environmental effects on genetic differentiation using bootstrap method, 3) interfaces to read and manipulate GENEPOP format data files and allele/haplotype frequency format files.

The Clutter model is a significant forest growth simulation tool. Grounded on individual trees and comprehensively considering factors such as competition among trees and the impact of environmental elements on growth, it can accurately reflect the growth process of forest stands. It can be applied in areas like forest resource management, harvesting planning, and ecological research. With the help of the Clutter model, people can better understand the dynamic changes of forests and provide a scientific basis for rational forest management and protecting the ecological environment. This R package can effectively realize the construction of forest growth and harvest models based on the Clutter model and achieve optimized forest management.References: Farias A, Soares C, Leite H et al(2021)<doi:10.1007/s10342-021-01380-1>. Guera O, Silva J, Ferreira R, et al(2019)<doi:10.1590/2179-8087.038117>.

Extends the fable framework to support forecasting methods specifically designed for intermittent time series data, where demand occurs sporadically with many zero values. All methods produce probabilistic forecasts returned as distributional objects. The returned forecasts can be used to evaluate accuracy, plot and print the results seamlessly with fable'. The methods include: Harvey, Fernandes (1989) <doi:10.1080/07350015.1989.10509750>, Willemain, Smart, Schwarz (2004) <doi:10.1016/S0169-2070(03)00013-X>, Zhou, Viswanathan (2011) <doi:10.1016/j.ijpe.2010.09.021>, Snyder, Ord, Beaumont (2012) <doi:10.1016/j.ijforecast.2011.03.009>, Kolassa (2016) <doi:10.1016/j.ijforecast.2015.12.004>, Hasni, Aguir, Babai, Jemai (2019) <doi:10.1080/00207543.2018.1424375>, Damato, Azzimonti, Corani (2025) <doi:10.1016/j.ijforecast.2025.10.001>, Sbrana (2025) <doi:10.1080/01605682.2025.2569661>.

This package provides a suite of methods for detecting influential subjects in longitudinal datasets, particularly when observations occur at irregular time points. The methods identify individuals whose response trajectories deviate significantly from the population pattern, enabling detection of anomalies or subjects exerting undue influence on model outcomes.

The FLEX method, developed by Yoon and Choi (2013) <doi:10.1007/978-3-642-33042-1_21>, performs least squares estimation for fuzzy predictors and outcomes, generating crisp regression coefficients by minimizing the distance between observed and predicted outcomes. It also provides functions for fuzzifying data and inference tasks, including significance testing, fit indices, and confidence interval estimation.

Perform Maximum Likelihood Factor analysis on a covariance matrix or data matrix.

Separates diffusive and ebullitive (bubble) fluxes from continuous concentration measurements using a running variance approach. Ebullitive events are identified when the running variance exceeds a user-set threshold. Diffusive fluxes are calculated via linear regression on the non-ebullitive portion of the data. See Sø et al. (2024) <doi:10.1029/2024JG008035> for details.

This package provides a handy tool to calculate carbon footprints from air travel based on three-letter International Air Transport Association (IATA) airport codes or latitude and longitude. footprint first calculates the great-circle distance between departure and arrival destinations. It then uses the Department of Environment, Food & Rural Affairs (DEFRA) greenhouse gas conversion factors for business air travel to estimate the carbon footprint. These conversion factors consider trip length, flight class (e.g. economy, business), and emissions metric (e.g. carbon dioxide equivalent, methane).

General linear modeling with multiple responses (MANCOVA). An overall p-value for each model term is calculated by the 50-50 MANOVA method by Langsrud (2002) <doi:10.1111/1467-9884.00320>, which handles collinear responses. Rotation testing, described by Langsrud (2005) <doi:10.1007/s11222-005-4789-5>, is used to compute adjusted single response p-values according to familywise error rates and false discovery rates (FDR). The approach to FDR is described in the appendix of Moen et al. (2005) <doi:10.1128/AEM.71.4.2086-2094.2005>. Unbalanced designs are handled by Type II sums of squares as argued in Langsrud (2003) <doi:10.1023/A:1023260610025>. Furthermore, the Type II philosophy is extended to continuous design variables as described in Langsrud et al. (2007) <doi:10.1080/02664760701594246>. This means that the method is invariant to scale changes and that common pitfalls are avoided.

This package provides a general estimation framework for multi-state Markov processes with flexible specification of the transition intensities. The log-transition intensities can be specified through Generalised Additive Models which allow for virtually any type of covariate effect. Elementary specifications such as time-homogeneous processes and simple parametric forms are also supported. There are no limitations on the type of process one can assume, with both forward and backward transitions allowed and virtually any number of states.

For ordinal rating data, consider the accelerated EM algorithm to estimate and test models within the family of CUB models (where CUB stands for Combination of a discrete Uniform and a shifted Binomial distributions). The procedure is built upon Louis identity for the observed information matrix. Best-subset variable selection is then implemented since it becomes more feasible from the computational point of view.

SHE, FORAM Index and ABC Method analyses and custom plot functions for community data.

This package provides four addons for analyzing trends and unit roots in financial time series: (i) functions for the density and probability of the augmented Dickey-Fuller Test, (ii) functions for the density and probability of MacKinnon's unit root test statistics, (iii) reimplementations for the ADF and MacKinnon Test, and (iv) an urca Unit Root Test Interface for Pfaff's unit root test suite.

Helps you imagine your data before you collect it. Hierarchical data structures and correlated data can be easily simulated, either from random number generators or by resampling from existing data sources. This package is faster with data.table and mvnfast installed.

Several functions to compute indicators for organization and efficiency in visual foraging, multi-target visual search, and cancellation tasks. The current version of this package includes the following indicators: best-r, mean Inter-target Distance, Percentage Above Optimal (PAO) scan path, and intersections in the scan path. For more detailed descriptions, see Mark et al. (2004) <doi:10.1212/01.WNL.0000131947.08670.D4>.

Allows prophet models from the prophet package to be used in a tidy workflow with the modelling interface of fabletools'. This extends prophet to provide enhanced model specification and management, performance evaluation methods, and model combination tools.