Discrimination Mitigation for Machine Learning Models

Based on different statistical definitions of discrimination, several methods have been proposed to detect and mitigate social inequality in machine learning models. This package aims to provide an alternative to fairness treatment in predictive models. The ROC method implemented in this package is described by Kamiran, Karim and Zhang (2012) < https://ieeexplore.ieee.org/document/6413831/>.

Computation and Plots of Influence Functions for Risk and Performance Measures

Computes the influence functions time series of the returns for the risk and performance measures as mentioned in Chen and Martin (2018) < https://www.ssrn.com/abstract=3085672>, as well as in Zhang et al. (2019) < https://www.ssrn.com/abstract=3415903>. Also evaluates estimators influence functions at a set of parameter values and plots them to display the shapes of the influence functions.

Bayesian Modeling via Frequentist Goodness-of-Fit

A Bayesian data modeling scheme that performs four interconnected tasks: (i) characterizes the uncertainty of the elicited parametric prior; (ii) provides exploratory diagnostic for checking prior-data conflict; (iii) computes the final statistical prior density estimate; and (iv) executes macro- and micro-inference. Primary reference is Mukhopadhyay, S. and Fletcher, D. 2018 paper "Generalized Empirical Bayes via Frequentist Goodness of Fit" (< https://www.nature.com/articles/s41598-018-28130-5 >).

Choi and Hall Style Data Sharpening

Functions for use in perturbing data prior to use of nonparametric smoothers and clustering.

Generalized Additive Latent and Mixed Models

Estimates generalized additive latent and mixed models using maximum marginal likelihood, as defined in Sorensen et al. (2023) , which is an extension of Rabe-Hesketh and Skrondal (2004)'s unifying framework for multilevel latent variable modeling . Efficient computation is done using sparse matrix methods, Laplace approximation, and automatic differentiation. The framework includes generalized multilevel models with heteroscedastic residuals, mixed response types, factor loadings, smoothing splines, crossed random effects, and combinations thereof. Syntax for model formulation is close to 'lme4' (Bates et al. (2015) ) and 'PLmixed' (Rockwood and Jeon (2019) ).

Data sets from "SAS System for Mixed Models"

Data sets and sample lmer analyses corresponding to the examples in Littell, Milliken, Stroup and Wolfinger (1996), "SAS System for Mixed Models", SAS Institute.

Geographic and Taxonomic Occurrence R-Based Scrubbing

Streamlines downloading and cleaning biodiversity data from Integrated Digitized Biocollections (iDigBio) and the Global Biodiversity Information Facility (GBIF).

Design of Experiments and Factorial Plans Utilities

A number of functions to create and analyze factorial plans according to the Design of Experiments (DoE) approach, with the addition of some utility function to perform some statistical analyses. DoE approach follows the approach in "Design and Analysis of Experiments" by Douglas C. Montgomery (2019, ISBN:978-1-119-49244-3). The package also provides utilities used in the course "Analysis of Data and Statistics" at the University of Trento, Italy.

Compute Decision Interval and Average Run Length for CUSUM Charts

Computation of decision intervals (H) and average run lengths (ARL) for CUSUM charts. Details of the method are seen in Hawkins and Olwell (2012): Cumulative sum charts and charting for quality improvement, Springer Science & Business Media.

Similarity-Based Segmentation of Multidimensional Signals

A dynamic programming solution to segmentation based on maximization of arbitrary similarity measures within segments. The general idea, theory and this implementation are described in Machne, Murray & Stadler (2017) . In addition to the core algorithm, the package provides time-series processing and clustering functions as described in the publication. These are generally applicable where a `k-means` clustering yields meaningful results, and have been specifically developed for clustering of the Discrete Fourier Transform of periodic gene expression data (`circadian' or `yeast metabolic oscillations'). This clustering approach is outlined in the supplemental material of Machne & Murray (2012) ), and here is used as a basis of segment similarity measures. Notably, the time-series processing and clustering functions can also be used as stand-alone tools, independent of segmentation, e.g., for transcriptome data already mapped to genes.