Estimation of the average treatment effect when controlling for high-dimensional confounders using debiased inverse propensity score weighting (DIPW). DIPW relies on the propensity score following a sparse logistic regression model, but the regression curves are not required to be estimable. Despite this, our package also allows the users to estimate the regression curves and take the estimated curves as input to our methods. Details of the methodology can be found in Yuhao Wang and Rajen D. Shah (2020) "Debiased Inverse Propensity Score Weighting for Estimation of Average Treatment Effects with High-Dimensional Confounders" <arXiv:2011.08661>. The package relies on the optimisation software MOSEK <https://www.mosek.com/> which must be installed separately; see the documentation for Rmosek'.

Fit response surfaces for datasets with latent-variable Gaussian process modeling, predict responses for new inputs, and plot latent variables locations in the latent space (only 1D or 2D). The input variables of the datasets can be quantitative, qualitative/categorical or mixed. The output variable of the datasets is a scalar (quantitative). The optimization of the likelihood function is done using a successive approximation/relaxation algorithm similar to another GP modeling package "GPM". The modeling method is published in "A Latent Variable Approach to Gaussian Process Modeling with Qualitative and Quantitative Factors" by Yichi Zhang, Siyu Tao, Wei Chen, and Daniel W. Apley (2018) <doi:10.48550/arXiv.1806.07504>. The package is developed in IDEAL of Northwestern University.

Computation of predictive information criteria (PIC) from select model object classes for model selection in predictive contexts. In contrast to the more widely used Akaike Information Criterion (AIC), which are derived under the assumption that target(s) of prediction (i.e. validation data) are independently and identically distributed to the fitting data, the PIC are derived under less restrictive assumptions and thus generalize AIC to the more practically relevant case of training/validation data heterogeneity. The methodology featured in this package is based on Flores (2021) <https://iro.uiowa.edu/esploro/outputs/doctoral/A-new-class-of-information-criteria/9984097169902771?institution=01IOWA_INST> "A new class of information criteria for improved prediction in the presence of training/validation data heterogeneity".

This package provides a novel meta-learning framework for forecast model selection using time series features. Many applications require a large number of time series to be forecast. Providing better forecasts for these time series is important in decision and policy making. We propose a classification framework which selects forecast models based on features calculated from the time series. We call this framework FFORMS (Feature-based FORecast Model Selection). FFORMS builds a mapping that relates the features of time series to the best forecast model using a random forest. seer package is the implementation of the FFORMS algorithm. For more details see our paper at <https://www.monash.edu/business/econometrics-and-business-statistics/research/publications/ebs/wp06-2018.pdf>.

This package provides a collection of procedures for analysing, visualising, and managing single-case data. Multi-phase and multi-baseline designs are supported. Analysing methods include regression models (multilevel, multivariate, bayesian), between case standardised mean difference, overlap indices ('PND', PEM', PAND', NAP', PET', tau-u', IRD', baseline corrected tau', CDC'), and randomization tests. Data preparation functions support outlier detection, handling missing values, scaling, and custom transformations. An export function helps to generate html, word, and latex tables in a publication friendly style. A shiny app allows to use scan in a graphical user interface. More details can be found in the online book Analyzing single-case data with R and scan', Juergen Wilbert (2026) <https://jazznbass.github.io/scan-Book/>.

In randomized controlled trial (RCT), balancing covariate is often one of the most important concern. CARM package provides functions to balance the covariates and generate allocation sequence by covariate-adjusted Adaptive Randomization via Mahalanobis-distance (ARM) for RCT. About what ARM is and how it works please see Y. Qin, Y. Li, W. Ma, H. Yang, and F. Hu (2024). "Adaptive randomization via Mahalanobis distance" Statistica Sinica. <doi:10.5705/ss.202020.0440>. In addition, the package is also suitable for the randomization process of multi-arm trials. For details, please see Yang H, Qin Y, Wang F, et al. (2023). "Balancing covariates in multi-arm trials via adaptive randomization" Computational Statistics & Data Analysis.<doi:10.1016/j.csda.2022.107642>.

This package provides a computationally efficient method (at least at the time of development!) for robust additive modeling based on the gamma divergence, implementing the methodology of Hui, Ding, and Sugasawa (2026) <doi:10.1007/s11222-026-10960-3>. The method assumes an identity link and normally distributed errors, and applies the gamma divergence to the resulting restricted maximum likelihood function so as to obtain a loss function that is less sensitive to outlying responses. Because gdam uses a model fitted via the mgcv package as the starting point for optimization, and leverages many of its existing techniques for estimation and inference, users can take advantage of many of the smoothing options available in the mgcv package for constructing additive models.

This package implements indirect demographic methods for estimating adult mortality from orphanhood data. The package includes the standard Brass and Hill (1973) method <https://scholar.google.com/scholar_lookup?&title=Estimating%20Adult%20Mortality%20from%20Orphanhood&pages=111-23&publication_year=1973&author=Brass%2CW.&author=Hill.%2CK.>, the regression-based approach developed by Timaeus (1992) <https://pubmed.ncbi.nlm.nih.gov/12317481/>, and the adjustments proposed by Luy (2012) <doi:10.1007/s13524-012-0101-4> for low-mortality populations. A relational model is used to harmonize estimates into comparable adult mortality indicators. The package also provides diagnostic tools to assess the sensitivity of results to assumptions about the mean age of childbearing and the choice of model life table family.

This package performs iterative proportional updating given a seed table and an arbitrary number of marginal distributions. This is commonly used in population synthesis, survey raking, matrix rebalancing, and other applications. For example, a household survey may be weighted to match the known distribution of households by size from the census. An origin/ destination trip matrix might be balanced to match traffic counts. The approach used by this package is based on a paper from Arizona State University (Ye, Xin, et. al. (2009) <http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.537.723&rep=rep1&type=pdf>). Some enhancements have been made to their work including primary and secondary target balance/importance, general marginal agreement, and weight restriction.

Imputes missing biomarker measurements in a wide longitudinal serum panel with gradient-boosted decision trees, groups the completed panel by Bayesian consensus clustering, and compares the resulting patient subgroups by Kaplan-Meier, log-rank and Cox analysis. The imputation learner is described in Ke et al. (2017) <https://papers.nips.cc/paper/6907-lightgbm-a-highly-efficient-gradient-boosting-decision-tree> and the clustering method in Lock and Dunson (2013) <doi:10.1093/bioinformatics/btt425>. Imputed values are conditional-mean predictions, so the procedure is a machine-learning single imputation; the completions carry no between-imputation variance and must not be pooled by Rubin's rules. Two panels from Gene Expression Omnibus accession GSE65622 are included, one for each survival endpoint.

The multiple instance data set consists of many independent subjects (called bags) and each subject is composed of several components (called instances). The outcomes of such data set are binary or categorical responses, and, we can only observe the subject-level outcomes. For example, in manufacturing processes, a subject is labeled as "defective" if at least one of its own components is defective, and otherwise, is labeled as "non-defective". The milr package focuses on the predictive model for the multiple instance data set with binary outcomes and performs the maximum likelihood estimation with the Expectation-Maximization algorithm under the framework of logistic regression. Moreover, the LASSO penalty is attached to the likelihood function for simultaneous parameter estimation and variable selection.

With the provision of several tools and templates the MOSAIC project (DFG-Grant Number HO 1937/2-1) supports the implementation of a central data management in epidemiological research projects. The MOQA package enables epidemiologists with none or low experience in R to generate basic data quality reports for a wide range of application scenarios. See <https://mosaic-greifswald.de/> for more information. Please read and cite the corresponding open access publication (using the former package-name) in METHODS OF INFORMATION IN MEDICINE by M. Bialke, H. Rau, T. Schwaneberg, R. Walk, T. Bahls and W. Hoffmann (2017) <doi:10.3414/ME16-01-0123>. <https://methods.schattauer.de/en/contents/most-recent-articles/issue/2483/issue/special/manuscript/27573/show.html>.

An updated and extended version of spm package, by introducing some further novel functions for modern statistical methods (i.e., generalised linear models, glmnet, generalised least squares), thin plate splines, support vector machine, kriging methods (i.e., simple kriging, universal kriging, block kriging, kriging with an external drift), and novel hybrid methods (228 hybrids plus numerous variants) of modern statistical methods or machine learning methods with mathematical and/or univariate geostatistical methods for spatial predictive modelling. For each method, two functions are provided, with one function for assessing the predictive errors and accuracy of the method based on cross-validation, and the other for generating spatial predictions. It also contains a couple of functions for data preparation and predictive accuracy assessment.

Interaction between a genetic variant (e.g., a single nucleotide polymorphism) and an environmental variable (e.g., physical activity) can have a shared effect on multiple phenotypes (e.g., blood lipids). We implement a two-step method to test for an overall interaction effect on multiple phenotypes. In first step, the method tests for an overall marginal genetic association between the genetic variant and the multivariate phenotype. The genetic variants which show an evidence of marginal overall genetic effect in the first step are prioritized while testing for an overall gene-environment interaction effect in the second step. Methodology is available from: A Majumdar, KS Burch, T Haldar, S Sankararaman, B Pasaniuc, WJ Gauderman, JS Witte (2020) <doi:10.1093/bioinformatics/btaa1083>.

Gradient boosting is a powerful statistical learning method known for its ability to model complex relationships between predictors and outcomes while performing inherent variable selection. However, traditional gradient boosting methods lack flexibility in handling longitudinal data where within-subject correlations play a critical role. In this package, we propose a novel approach Mixed Effect Gradient Boosting ('MEGB'), designed specifically for high-dimensional longitudinal data. MEGB incorporates a flexible semi-parametric model that embeds random effects within the gradient boosting framework, allowing it to account for within-individual covariance over time. Additionally, the method efficiently handles scenarios where the number of predictors greatly exceeds the number of observations (p>>n) making it particularly suitable for genomics data and other large-scale biomedical studies.

An implementation of the surrogate approach to residuals and diagnostics for ordinal and general regression models; for details, see Liu and Zhang (2018) <doi:10.1080/01621459.2017.1292915> and Greenwell et al. (2018) <doi:10.32614/RJ-2018-004>. These residuals can be used to construct standard residual plots for model diagnostics (e.g., residual-vs-fitted value plots, residual-vs-covariate plots, Q-Q plots, etc.). The package also provides plot methods for producing standard diagnostic plots using lightweight base R graphics (via tinyplot'). The package currently supports cumulative link models from packages MASS', ordinal', rms', and VGAM', binary and other generalized linear models fit with glm', and multinomial models or general classifiers via a matrix of predicted class probabilities.

piRNAs (short for PIWI-interacting RNAs) and their PIWI protein partners play a key role in fertility and maintaining genome integrity by restricting mobile genetic elements (transposons) in germ cells. piRNAs originate from genomic regions known as piRNA clusters. The piRNA Cluster Builder (PICB) is a versatile toolkit designed to identify genomic regions with a high density of piRNAs. It constructs piRNA clusters through a stepwise integration of unique and multimapping piRNAs and offers wide-ranging parameter settings, supported by an optimization function that allows users to test different parameter combinations to tailor the analysis to their specific piRNA system. The output includes extensive metadata columns, enabling researchers to rank clusters and extract cluster characteristics.

Package contains functions for analyzing check-all-that-apply (CATA) data from consumer and sensory tests. Cochran's Q test, McNemar's test, and Penalty-Lift analysis are provided; for details, see Meyners, Castura & Carr (2013) <doi:10.1016/j.foodqual.2013.06.010>. Cluster analysis can be performed using b-cluster analysis, then evaluated using various measures; for details, see Castura, Meyners, Varela & Næs (2022) <doi:10.1016/j.foodqual.2022.104564>. Consumers can also be clustered on their product-related hedonic responses; see Castura, Meyners, Pohjanheimo, Varela & Næs (2023) <doi:10.1111/joss.12860>. Permutation tests based on the L1-norm methods are provided; for details, see Chaya, Castura & Greenacre (2025) <doi:10.1016/j.foodqual.2025.105639>.

Supplements for a book, "iTOS" = "Introduction to the Theory of Observational Studies." Data sets are aHDL from Rosenbaum (2023a) <doi:10.1111/biom.13558> and bingeM from Rosenbaum (2023b) <doi:10.1111/biom.13921>. The function makematch() uses two-criteria matching from Zhang et al. (2023) <doi:10.1080/01621459.2021.1981337> to create the matched data bingeM from binge'. The makematch() function also implements optimal matching (Rosenbaum (1989) <doi:10.2307/2290079>) and matching with fine or near-fine balance (Rosenbaum et al. (2007) <doi:10.1198/016214506000001059> and Yang et al (2012) <doi:10.1111/j.1541-0420.2011.01691.x>). The book makes use of two other R packages, weightedRank and tightenBlock'.

Plug-in and difference-based long-run covariance matrix estimation for time series regression. Two applications of hypothesis testing are also provided. The first one is for testing for structural stability in coefficient functions. The second one is aimed at detecting long memory in time series regression. Lujia Bai and Weichi Wu (2024)<doi:10.3150/23-BEJ1680> Zhou Zhou and Wei Biao Wu(2010)<doi:10.1111/j.1467-9868.2010.00743.x> Jianqing Fan and Wenyang Zhang<doi:10.1214/aos/1017939139> Lujia Bai and Weichi Wu(2024)<doi:10.1093/biomet/asae013> Dimitris N. Politis, Joseph P. Romano, Michael Wolf(1999)<doi:10.1007/978-1-4612-1554-7> Weichi Wu and Zhou Zhou(2018)<doi:10.1214/17-AOS1582>.

Fit penalized splines mixed-effects models (a special case of additive models) for large longitudinal datasets. The package includes a psme() function that (1) relies on package mgcv for constructing population and subject smooth functions as penalized splines, (2) transforms the constructed additive model to a linear mixed-effects model, (3) exploits package lme4 for model estimation and (4) backtransforms the estimated linear mixed-effects model to the additive model for interpretation and visualization. See Pedersen et al. (2019) <doi:10.7717/peerj.6876> and Bates et al. (2015) <doi:10.18637/jss.v067.i01> for an introduction. Unlike the gamm() function in mgcv', the psme() function is fast and memory-efficient, able to handle datasets with millions of observations.

Training of neural networks for classification and regression tasks using mini-batch gradient descent. Special features include a function for training autoencoders, which can be used to detect anomalies, and some related plotting functions. Multiple activation functions are supported, including tanh, relu, step and ramp. For the use of the step and ramp activation functions in detecting anomalies using autoencoders, see Hawkins et al. (2002) <doi:10.1007/3-540-46145-0_17>. Furthermore, several loss functions are supported, including robust ones such as Huber and pseudo-Huber loss, as well as L1 and L2 regularization. The possible options for optimization algorithms are RMSprop, Adam and SGD with momentum. The package contains a vectorized C++ implementation that facilitates fast training through mini-batch learning.

Mixed models for repeated measures (MMRM) are a popular choice for analyzing longitudinal continuous outcomes in randomized clinical trials and beyond; see Cnaan, Laird and Slasor (1997) <doi:10.1002/(SICI)1097-0258(19971030)16:20%3C2349::AID-SIM667%3E3.0.CO;2-E> for a tutorial and Mallinckrodt, Lane, Schnell, Peng and Mancuso (2008) <doi:10.1177/009286150804200402> for a review. This package implements MMRM based on the marginal linear model without random effects using Template Model Builder ('TMB') which enables fast and robust model fitting. Users can specify a variety of covariance matrices, weight observations, fit models with restricted or standard maximum likelihood inference, perform hypothesis testing with Satterthwaite or Kenward-Roger adjustment, and extract least square means estimates by using emmeans'.

Parameter estimation and classification for Gaussian Mixture Models (GMMs) in the presence of missing data. This package complements existing implementations by allowing for both missing elements in the input vectors and full (as opposed to strictly diagonal) covariance matrices. Estimation is performed using an expectation conditional maximization algorithm that accounts for missingness of both the cluster assignments and the vector components. The output includes the marginal cluster membership probabilities; the mean and covariance of each cluster; the posterior probabilities of cluster membership; and a completed version of the input data, with missing values imputed to their posterior expectations. For additional details, please see McCaw ZR, Julienne H, Aschard H. "Fitting Gaussian mixture models on incomplete data." <doi:10.1186/s12859-022-04740-9>.