One hundred fifty-six new packages made it to CRAN in October. Here are my “Top 40” selections in eight categories: Computational Methods, Data, Epidemiology, Mathematics, Machine Learning, Statistics, Utilities, and Visualization.

Computational Methods

mcmcensemble v 2.0: Provides ensemble samplers for affine-invariant Monte Carlo Markov Chain algorithms which allow a faster convergence for badly scaled estimation problems. Two samplers are included: the ‘differential.evolution’ sampler from the Braak and Vrugt (2008) and the ‘stretch’ sampler from Goodman and Weare (2010). Look here for examples.

psqn v0.1.3: Provides quasi-Newton methods to minimize partially separable functions. The methods are described by Nocedal and Wright (2006). There is an Introduction and a vignette on the Partially Separable Quasi-Newton method.

Data

AirSensor v1.0.2: Allows processing and displaying data from air quality sensors. Initial focus is on PM2.5 measurements from sensors produced by PurpleAir.

fflr v0.3.12: Provides functions to format the raw data from the ESPN fantasy football API into tidy tables. See the vignette.

pmlbr v0.2.0: Provides access to more than 150 classification and regression data sets in the University of Pennsylvania’s PMLB repository. See README to get started.

podr v0.0.5: Allows users to to connect, access and review over 250 datasets in Pharmaceutical User Software Exchange (PHUSE) open data repository (PODR). See the vignette for details.

starwarsdb v0.1.2: Provides the data from the Star Wars API as a set of relational tables or a DuckDB database. Look here for an example.

Epidemiology

anovir v0.1.0: Implements maximum likelihood techniques to estimate virulence in population dynamics models. See the pre-print and the eleven vignettes including Introduction, Confidence intervals and Worked examples I and II.

epifitter v0.1.0: Provides functions for fitting two-parameter population dynamics models to proportion data for single or multiple epidemics using either linear or non-linear regression. See Madden et al. (2007) for background and the vignettes on fitting models and simulating disease progress.

epitweetr v0.1.24: Allows users to automatically monitor trends of tweets by time, place and topic aiming at detecting public health threats early through the detection of signals (e.g. an unusual increase in the number of tweets). It was designed to focus on infectious diseases, but can be adapted to other applications by modifying the topics and keywords. See the vignette.

i2extras v0.0.2: Provides functions to work with ‘incidence2’ objects, including a simplified interface for trend fitting and peak estimation. This package is part of the RECON toolkit for outbreak analysis. There are vignettes on Fitting epicurves and Peak Estimation.

IBMPopSim v0.3.1: Provides functions to simulate the random evolution of structured population dynamics, called stochastic Individual Based Models (IBMs). Users can simulate the random evolution of a population in which individuals are characterized by their date of birth, a set of attributes, and their potential date of death. See Ferrière and Tran (2009) and Bansaye and Méléard (2015) for background. There is a package overview and vignettes on C++ essentials, Human populations, Human populations with changing characteristics, Insurance portfolio, and populations with genetically variable traits.

MGDrivE2 v1.0.1: Provides a simulation modeling framework which significantly extends capabilities of the MGDrivE package with a new mathematical and computational framework based on stochastic Petri nets. To get started with this package see the vignettes on SEIR dynamics, Meta population dynamics, One node dynamics, Inhomogenous stochastic processes, Life-cycle dynamics, One node lifecycle dynamics, data analysis, and advanced topics

msce v1.0.1: Provide functions to calculate hazard and survival functions for multi-stage clonal expansion models used in cancer epidemiology. There is a vignette on fitting incidence data.

Machine Learning

acumos v0.4-1: Provides access to the Linux Foundation Acumos open source framework intended to make it easy to build, share, and deploy AI apps. Look here for information on the Acumos R CLient.

bigSurvSGD v0.0.1: Provides a function to fit Cox models via stochastic gradient descent which avoids the computational instability of the standard Cox Model. Functions scales up with large datasets and accommodate datasets that do not fit the memory. See Aliasghar et al. (2020) for details.

deepredeff v0.1.0: Implements a tool that contains trained deep learning models for predicting effector proteins using a set of known experimentally validated effectors from either bacteria, fungi, or oomycetes. Kristianingsih and MacLean (2020) for background, and the overview and vignette on prediction.

MKclass v0.3: Implements performance measures and scores for statistical classification including accuracy, sensitivity, specificity, recall, similarity coefficients, AUC, GINI index, Brier score and more. It calculates optimal cut-offs and decisions stumps according to (Iba and Langley (1991) follows Lemeshow and Hosmer and Hosmer et al. (1997) for goodness of fit tests and Porta (2014) for epidemiological risk measures. See the vignette to get started.

mlr3hyperband v0.1.0: Implements the bandit-based hyperparameter optimization method of Li et al. (2016). Look here for an example.

Mathematics

tdaunif Provides functions to randomly sample from simple manifolds. See Arvo (1995) and Diaconis, Holmes, and Shahshahani (2013) for the theory, and the vignette to get started.

Science

SoundShape v1.0: Implements the eignensound method of MacLeod et al. (2013) to compare stereotyped sound from different species. The vignette will make you want to start comparing sound waves.

Statistics

anscombiser v1.0.0: Provides functions to create data sets with identical summary statistics: i.e. identical marginal sample means and sample variances, sample correlation, least squares regression coefficients and coefficient of determination, that that look amusingly different. See the vignette for examples.

gglm v0.1.0: Extends ggplot2 for creating common diagnostic plots associated with linear models. Look here for examples.

microcluster v0.1.0: Implements the method in Betancourt et al. (2020) to perform microclustering models for categorical data. The vignette offers an example.

PressPurt v1.0.2: Provides functions to identify the most sensitive interactions within a network which can be described by differential equations, in order to produce qualitatively robust predictions to a press perturbation. See Koslicki & Novak (2017) for background. There a tutorial and a vignette for set up.

quantdr v1.0.0: Provides functions to perform dimension reduction for conditional quantiles by determining the directions that span the central quantile subspace (CQS). The vignette contains the details.

singcar v0.1.1: Implements frequentist and Bayesian methods to compare single cases to control populations. See Crawford and Howell (1998) and Crawford and Garthwaite (2005) for background, and the vignette for an introduction.

statgenGxE v1.0.3: Provides functions to facilitate the analysis of multi-environment data from plant breeding experiments following the analyses described in Malosetti et al. (2013). See the vignette for examples.

survivalMPLdc v0.1.1: Implements functions to fit Cox proportional hazard models under dependent right censoring using copula and maximum penalized likelihood methods. See the vignette for examples.

Time Series

FoReco v0.1.1: Provides bottom-up, optimal and heuristic combination forecast reconciliation procedures for cross-sectional, temporal, and cross-temporal linearly constrained time series. There is an introduction and another vignette on average relative accuracy indices.

modeltime.ensemble v0.3.0: Implements time series ensemble forecasting methods including model averaging, weighted averaging, and stacking. See Pavlyshenko (2019) for the theory and the vignette for an introduction.

Utilities

findInFiles v0.1.2: Enables users to search for a pattern in a folder and display the results in the RStudio viewer pane or in an R Markdown file or Shiny app.

groundhog v1.0.0: Assists with reproducibility by providing functions for version specific package loading. See the vignette.

monaco v0.1.0: Implements an HTML widget rendering of the Monaco editor which is useful for JavaScript. Look here for examples.

oskeyring v0.1.1: Aims to support all features of the system credential store, including non-portable ones. It supports Keychain on macOS, and Credential Manager on Windows. See the keyring package if you need a portable API. README describes how to get started.

PDE v1.1.1: Enables extracting information and tables from pdf files based on search words. The vignette has several examples.

representr v0.1.1: Implements the record linkage methods of Kaplan et al. (2020) to create representative records for use in downstream tasks after entity resolution is performed. See the vignette.

Visualizations

imbibe v0.1.0: Provides a set of fast, chainable image-processing operations which are applicable to images of two, three or four dimensions, particularly medical images. Look here for an example.

r3dmol v0.1.0: Provides functions to create and manipulate rich and fully interactive 3D visualizations of molecular data that can be included in Shiny apps and R markdown documents. See the vignette for an introduction.

visualpred v0.1.0: Provides 2D point and contour plots for visualizing binary classification models. There is an introduction and vignettes on plotting outliers, comparing algorithms, and advanced settings.