Allows users to create weighted confusion matrices and accuracy metrics that help with the model selection process for classification problems, where distance from the correct category is important. The package includes several weighting schemes which can be parameterized, as well as custom configuration options. Furthermore, users can decide whether they wish to positively or negatively affect the accuracy score as a result of applying weights to the confusion matrix. Functions are included to calculate accuracy metrics for imbalanced data. Finally, wconf integrates well with the caret package, but it can also work standalone when provided data in matrix form. References: Kuhn, M. (2008) "Building Perspective Models in R Using the caret Package" <doi:10.18637/jss.v028.i05> Monahov, A. (2021) "Model Evaluation with Weighted Threshold Optimization (and the mewto R package)" <doi:10.2139/ssrn.3805911> Monahov, A. (2024) "Improved Accuracy Metrics for Classification with Imbalanced Data and Where Distance from the Truth Matters, with the wconf R Package" <doi:10.2139/ssrn.4802336> Starovoitov, V., Golub, Y. (2020). New Function for Estimating Imbalanced Data Classification Results. Pattern Recognition and Image Analysis, 295â 302 Van de Velden, M., Iodice D'Enza, A., Markos, A., Cavicchia, C. (2023) "A general framework for implementing distances for categorical variables" <doi:10.48550/arXiv.2301.02190>.

High-dimensional matrix factor models have drawn much attention in view of the fact that observations are usually well structured to be an array such as in macroeconomics and finance. In addition, data often exhibit heavy-tails and thus it is also important to develop robust procedures. We aim to address this issue by replacing the least square loss with Huber loss function. We propose two algorithms to do robust factor analysis by considering the Huber loss. One is based on minimizing the Huber loss of the idiosyncratic error's Frobenius norm, which leads to a weighted iterative projection approach to compute and learn the parameters and thereby named as Robust-Matrix-Factor-Analysis (RMFA), see the details in He et al. (2023)<doi:10.1080/07350015.2023.2191676>. The other one is based on minimizing the element-wise Huber loss, which can be solved by an iterative Huber regression algorithm (IHR), see the details in He et al. (2023) <arXiv:2306.03317>. In this package, we also provide the algorithm for alpha-PCA by Chen & Fan (2021) <doi:10.1080/01621459.2021.1970569>, the Projected estimation (PE) method by Yu et al. (2022)<doi:10.1016/j.jeconom.2021.04.001>. In addition, the methods for determining the pair of factor numbers are also given.

This package performs parametric and non-parametric estimation and simulation of drifting semi-Markov processes. The definition of parametric and non-parametric model specifications is also possible. Furthermore, three different types of drifting semi-Markov models are considered. These models differ in the number of transition matrices and sojourn time distributions used for the computation of a number of semi-Markov kernels, which in turn characterize the drifting semi-Markov kernel. For the parametric model estimation and specification, several discrete distributions are considered for the sojourn times: Uniform, Poisson, Geometric, Discrete Weibull and Negative Binomial. The non-parametric model specification makes no assumptions about the shape of the sojourn time distributions. Semi-Markov models are described in: Barbu, V.S., Limnios, N. (2008) <doi:10.1007/978-0-387-73173-5>. Drifting Markov models are described in: Vergne, N. (2008) <doi:10.2202/1544-6115.1326>. Reliability indicators of Drifting Markov models are described in: Barbu, V. S., Vergne, N. (2019) <doi:10.1007/s11009-018-9682-8>. We acknowledge the DATALAB Project <https://lmrs-num.math.cnrs.fr/projet-datalab.html> (financed by the European Union with the European Regional Development fund (ERDF) and by the Normandy Region) and the HSMM-INCA Project (financed by the French Agence Nationale de la Recherche (ANR) under grant ANR-21-CE40-0005).

This package implements the SE-test for equivalence according to Hoffelder et al. (2015) <DOI:10.1080/10543406.2014.920344>. The SE-test for equivalence is a multivariate two-sample equivalence test. Distance measure of the test is the sum of standardized differences between the expected values or in other words: the sum of effect sizes (SE) of all components of the two multivariate samples. The test is an asymptotically valid test for normally distributed data (see Hoffelder et al.,2015). The function SE.EQ() implements the SE-test for equivalence according to Hoffelder et al. (2015). The function SE.EQ.dissolution.profiles() implements a variant of the SE-test for equivalence for similarity analyses of dissolution profiles as mentioned in Suarez-Sharp et al.(2020) <DOI:10.1208/s12248-020-00458-9>). The equivalence margin used in SE.EQ.dissolution.profiles() is analogically defined as for the T2EQ approach according to Hoffelder (2019) <DOI:10.1002/bimj.201700257>) by means of a systematic shift in location of 10 [\% of label claim] of both dissolution profile populations. SE.EQ.dissolution.profiles() checks whether the weighted mean of the differences of the expected values of both dissolution profile populations is statistically significantly smaller than 10 [\% of label claim]. The weights are built up by the inverse variances.

The marginal treatment effect was introduced by Heckman and Vytlacil (2005) <doi:10.1111/j.1468-0262.2005.00594.x> to provide a choice-theoretic interpretation to instrumental variables models that maintain the monotonicity condition of Imbens and Angrist (1994) <doi:10.2307/2951620>. This interpretation can be used to extrapolate from the compliers to estimate treatment effects for other subpopulations. This package provides a flexible set of methods for conducting this extrapolation. It allows for parametric or nonparametric sieve estimation, and allows the user to maintain shape restrictions such as monotonicity. The package operates in the general framework developed by Mogstad, Santos and Torgovitsky (2018) <doi:10.3982/ECTA15463>, and accommodates either point identification or partial identification (bounds). In the partially identified case, bounds are computed using either linear programming or quadratically constrained quadratic programming. Support for four solvers is provided. Gurobi and the Gurobi R API can be obtained from <http://www.gurobi.com/index>. CPLEX can be obtained from <https://www.ibm.com/analytics/cplex-optimizer>. CPLEX R APIs Rcplex and cplexAPI are available from CRAN. MOSEK and the MOSEK R API can be obtained from <https://www.mosek.com/>. The lp_solve library is freely available from <http://lpsolve.sourceforge.net/5.5/>, and is included when installing its API lpSolveAPI', which is available from CRAN.

This package provides a suite of functions for performing mediation analysis with high-dimensional mediators. In addition to centralizing code from several existing packages for high-dimensional mediation analysis, we provide organized, well-documented functions for a handle of methods which, though programmed their original authors, have not previously been formalized into R packages or been made presentable for public use. The methods we include cover a broad array of approaches and objectives, and are described in detail by both our companion manuscript---"Methods for Mediation Analysis with High-Dimensional DNA Methylation Data: Possible Choices and Comparison"---and the original publications that proposed them. The specific methods offered by our package include the Bayesian sparse linear mixed model (BSLMM) by Song et al. (2019); high-dimensional mediation analysis (HDMA) by Gao et al. (2019); high-dimensional multivariate mediation (HDMM) by Chén et al. (2018); high-dimensional linear mediation analysis (HILMA) by Zhou et al. (2020); high-dimensional mediation analysis (HIMA) by Zhang et al. (2016); latent variable mediation analysis (LVMA) by Derkach et al. (2019); mediation by fixed-effect model (MedFix) by Zhang (2021); pathway LASSO by Zhao & Luo (2022); principal component mediation analysis (PCMA) by Huang & Pan (2016); and sparse principal component mediation analysis (SPCMA) by Zhao et al. (2020). Citations for the corresponding papers can be found in their respective functions.

Functionality for spatio-temporal modeling of large data sets is provided. A Gaussian process in space and time is defined through a stochastic partial differential equation (SPDE). The SPDE is solved in the spectral space, and after discretizing in time and space, a linear Gaussian state space model is obtained. When doing inference, the main computational difficulty consists in evaluating the likelihood and in sampling from the full conditional of the spectral coefficients, or equivalently, the latent space-time process. In comparison to the traditional approach of using a spatio-temporal covariance function, the spectral SPDE approach is computationally advantageous. See Sigrist, Kuensch, and Stahel (2015) <doi:10.1111/rssb.12061> for more information on the methodology. This package aims at providing tools for two different modeling approaches. First, the SPDE based spatio-temporal model can be used as a component in a customized hierarchical Bayesian model (HBM). The functions of the package then provide parameterizations of the process part of the model as well as computationally efficient algorithms needed for doing inference with the HBM. Alternatively, the adaptive MCMC algorithm implemented in the package can be used as an algorithm for doing inference without any additional modeling. The MCMC algorithm supports data that follow a Gaussian or a censored distribution with point mass at zero. Covariates can be included in the model through a regression term.

Implementation of a collection of MCMC methods for Bayesian structure learning of directed acyclic graphs (DAGs), both from continuous and discrete data. For efficient inference on larger DAGs, the space of DAGs is pruned according to the data. To filter the search space, the algorithm employs a hybrid approach, combining constraint-based learning with search and score. A reduced search space is initially defined on the basis of a skeleton obtained by means of the PC-algorithm, and then iteratively improved with search and score. Search and score is then performed following two approaches: Order MCMC, or Partition MCMC. The BGe score is implemented for continuous data and the BDe score is implemented for binary data or categorical data. The algorithms may provide the maximum a posteriori (MAP) graph or a sample (a collection of DAGs) from the posterior distribution given the data. All algorithms are also applicable for structure learning and sampling for dynamic Bayesian networks. References: J. Kuipers, P. Suter, G. Moffa (2022) <doi:10.1080/10618600.2021.2020127>, N. Friedman and D. Koller (2003) <doi:10.1023/A:1020249912095>, J. Kuipers and G. Moffa (2017) <doi:10.1080/01621459.2015.1133426>, M. Kalisch et al. (2012) <doi:10.18637/jss.v047.i11>, D. Geiger and D. Heckerman (2002) <doi:10.1214/aos/1035844981>, P. Suter, J. Kuipers, G. Moffa, N.Beerenwinkel (2023) <doi:10.18637/jss.v105.i09>.

This package performs frequentist inference for the extremal index of a stationary time series. Two types of methodology are used. One type is based on a model that relates the distribution of block maxima to the marginal distribution of series and leads to the semiparametric maxima estimators described in Northrop (2015) <doi:10.1007/s10687-015-0221-5> and Berghaus and Bucher (2018) <doi:10.1214/17-AOS1621>. Sliding block maxima are used to increase precision of estimation. A graphical block size diagnostic is provided. The other type of methodology uses a model for the distribution of threshold inter-exceedance times (Ferro and Segers (2003) <doi:10.1111/1467-9868.00401>). Three versions of this type of approach are provided: the iterated weight least squares approach of Suveges (2007) <doi:10.1007/s10687-007-0034-2>, the K-gaps model of Suveges and Davison (2010) <doi:10.1214/09-AOAS292> and a similar approach of Holesovsky and Fusek (2020) <doi:10.1007/s10687-020-00374-3> that we refer to as D-gaps. For the K-gaps and D-gaps models this package allows missing values in the data, can accommodate independent subsets of data, such as monthly or seasonal time series from different years, and can incorporate information from right-censored inter-exceedance times. Graphical diagnostics for the threshold level and the respective tuning parameters K and D are provided.

This package provides functions supporting the reading and parsing of internal e-book content from EPUB files. The epubr package provides functions supporting the reading and parsing of internal e-book content from EPUB files. E-book metadata and text content are parsed separately and joined together in a tidy, nested tibble data frame. E-book formatting is not completely standardized across all literature. It can be challenging to curate parsed e-book content across an arbitrary collection of e-books perfectly and in completely general form, to yield a singular, consistently formatted output. Many EPUB files do not even contain all the same pieces of information in their respective metadata. EPUB file parsing functionality in this package is intended for relatively general application to arbitrary EPUB e-books. However, poorly formatted e-books or e-books with highly uncommon formatting may not work with this package. There may even be cases where an EPUB file has DRM or some other property that makes it impossible to read with epubr'. Text is read as is for the most part. The only nominal changes are minor substitutions, for example curly quotes changed to straight quotes. Substantive changes are expected to be performed subsequently by the user as part of their text analysis. Additional text cleaning can be performed at the user's discretion, such as with functions from packages like tm or qdap'.

Fresh biomass determination is the key to evaluating crop genotypes response to diverse input and stress conditions and forms the basis for calculating net primary production. However, as conventional phenotyping approaches for measuring fresh biomass is time-consuming, laborious and destructive, image-based phenotyping methods are being widely used now. In the image-based approach, the fresh weight of the above-ground part of the plant depends on the projected area. For determining the projected area, the visual image of the plant is converted into the grayscale image by simply averaging the Red(R), Green (G) and Blue (B) pixel values. Grayscale image is then converted into a binary image using Otsuâ s thresholding method Otsu, N. (1979) <doi:10.1109/TSMC.1979.4310076> to separate plant area from the background (image segmentation). The segmentation process was accomplished by selecting the pixels with values over the threshold value belonging to the plant region and other pixels to the background region. The resulting binary image consists of white and black pixels representing the plant and background regions. Finally, the number of pixels inside the plant region was counted and converted to square centimetres (cm2) using the reference object (any object whose actual area is known previously) to get the projected area. After that, the projected area is used as input to the machine learning model (Linear Model, Artificial Neural Network, and Support Vector Regression) to determine the plant's fresh weight.

This package performs stability analysis of multi-environment trial data using parametric and non-parametric methods. Parametric methods includes Additive Main Effects and Multiplicative Interaction (AMMI) analysis by Gauch (2013) <doi:10.2135/cropsci2013.04.0241>, Ecovalence by Wricke (1965), Genotype plus Genotype-Environment (GGE) biplot analysis by Yan & Kang (2003) <doi:10.1201/9781420040371>, geometric adaptability index by Mohammadi & Amri (2008) <doi:10.1007/s10681-007-9600-6>, joint regression analysis by Eberhart & Russel (1966) <doi:10.2135/cropsci1966.0011183X000600010011x>, genotypic confidence index by Annicchiarico (1992), Murakami & Cruz's (2004) method, power law residuals (POLAR) statistics by Doring et al. (2015) <doi:10.1016/j.fcr.2015.08.005>, scale-adjusted coefficient of variation by Doring & Reckling (2018) <doi:10.1016/j.eja.2018.06.007>, stability variance by Shukla (1972) <doi:10.1038/hdy.1972.87>, weighted average of absolute scores by Olivoto et al. (2019a) <doi:10.2134/agronj2019.03.0220>, and multi-trait stability index by Olivoto et al. (2019b) <doi:10.2134/agronj2019.03.0221>. Non-parametric methods includes superiority index by Lin & Binns (1988) <doi:10.4141/cjps88-018>, nonparametric measures of phenotypic stability by Huehn (1990) <doi:10.1007/BF00024241>, TOP third statistic by Fox et al. (1990) <doi:10.1007/BF00040364>. Functions for computing biometrical analysis such as path analysis, canonical correlation, partial correlation, clustering analysis, and tools for inspecting, manipulating, summarizing and plotting typical multi-environment trial data are also provided.

Package provides functions for estimation and inference in Bayesian quantile regression with ordinal outcomes. An ordinal model with 3 or more outcomes (labeled OR1 model) is estimated by a combination of Gibbs sampling and Metropolis-Hastings (MH) algorithm. Whereas an ordinal model with exactly 3 outcomes (labeled OR2 model) is estimated using a Gibbs sampling algorithm. The summary output presents the posterior mean, posterior standard deviation, 95% credible intervals, and the inefficiency factors along with the two model comparison measures â logarithm of marginal likelihood and the deviance information criterion (DIC). The package also provides functions for computing the covariate effects and other functions that aids either the estimation or inference in quantile ordinal models. Rahman, M. A. (2016).â Bayesian Quantile Regression for Ordinal Models.â Bayesian Analysis, 11(1): 1-24 <doi: 10.1214/15-BA939>. Yu, K., and Moyeed, R. A. (2001). â Bayesian Quantile Regression.â Statistics and Probability Letters, 54(4): 437â 447 <doi: 10.1016/S0167-7152(01)00124-9>. Koenker, R., and Bassett, G. (1978).â Regression Quantiles.â Econometrica, 46(1): 33-50 <doi: 10.2307/1913643>. Chib, S. (1995). â Marginal likelihood from the Gibbs output.â Journal of the American Statistical Association, 90(432):1313â 1321, 1995. <doi: 10.1080/01621459.1995.10476635>. Chib, S., and Jeliazkov, I. (2001). â Marginal likelihood from the Metropolis-Hastings output.â Journal of the American Statistical Association, 96(453):270â 281, 2001. <doi: 10.1198/016214501750332848>.

In various domains, many datasets exhibit both high variable dependency and group structures, which necessitates their simultaneous estimation. This package provides functions for two subgroup identification methods based on penalized functions, both of which utilize factor model structures to adapt to data with cross-sectional dependency. The first method is the Subgroup Identification with Latent Factor Structure Method (SILFSM) we proposed. By employing Center-Augmented Regularization and factor structures, the SILFSM effectively eliminates data dependencies while identifying subgroups within datasets. For this model, we offer optimization functions based on two different methods: Coordinate Descent and our newly developed Difference of Convex-Alternating Direction Method of Multipliers (DC-ADMM) algorithms; the latter can be applied to cases where the distance function in Center-Augmented Regularization takes L1 and L2 forms. The other method is the Factor-Adjusted Pairwise Fusion Penalty (FA-PFP) model, which incorporates factor augmentation into the Pairwise Fusion Penalty (PFP) developed by Ma, S. and Huang, J. (2017) <doi:10.1080/01621459.2016.1148039>. Additionally, we provide a function for the Standard CAR (S-CAR) method, which does not consider the dependency and is for comparative analysis with other approaches. Furthermore, functions based on the Bayesian Information Criterion (BIC) of the SILFSM and the FA-PFP method are also included in SILFS for selecting tuning parameters. For more details of Subgroup Identification with Latent Factor Structure Method, please refer to He et al. (2024) <doi:10.48550/arXiv.2407.00882>.

In many binary classification applications, such as disease diagnosis and spam detection, practitioners commonly face the need to limit type I error (i.e., the conditional probability of misclassifying a class 0 observation as class 1) so that it remains below a desired threshold. To address this need, the Neyman-Pearson (NP) classification paradigm is a natural choice; it minimizes type II error (i.e., the conditional probability of misclassifying a class 1 observation as class 0) while enforcing an upper bound, alpha, on the type I error. Although the NP paradigm has a century-long history in hypothesis testing, it has not been well recognized and implemented in classification schemes. Common practices that directly limit the empirical type I error to no more than alpha do not satisfy the type I error control objective because the resulting classifiers are still likely to have type I errors much larger than alpha. As a result, the NP paradigm has not been properly implemented for many classification scenarios in practice. In this work, we develop the first umbrella algorithm that implements the NP paradigm for all scoring-type classification methods, including popular methods such as logistic regression, support vector machines and random forests. Powered by this umbrella algorithm, we propose a novel graphical tool for NP classification methods: NP receiver operating characteristic (NP-ROC) bands, motivated by the popular receiver operating characteristic (ROC) curves. NP-ROC bands will help choose in a data adaptive way and compare different NP classifiers.

Collection of R functions to do purely presence-only species distribution modeling with isolation forest (iForest) and its variations such as Extended isolation forest and SCiForest. See the details of these methods in references: Liu, F.T., Ting, K.M. and Zhou, Z.H. (2008) <doi:10.1109/ICDM.2008.17>, Hariri, S., Kind, M.C. and Brunner, R.J. (2019) <doi:10.1109/TKDE.2019.2947676>, Liu, F.T., Ting, K.M. and Zhou, Z.H. (2010) <doi:10.1007/978-3-642-15883-4_18>, Guha, S., Mishra, N., Roy, G. and Schrijvers, O. (2016) <https://proceedings.mlr.press/v48/guha16.html>, Cortes, D. (2021) <arXiv:2110.13402>. Additionally, Shapley values are used to explain model inputs and outputs. See details in references: Shapley, L.S. (1953) <doi:10.1515/9781400881970-018>, Lundberg, S.M. and Lee, S.I. (2017) <https://dl.acm.org/doi/abs/10.5555/3295222.3295230>, Molnar, C. (2020) <ISBN:978-0-244-76852-2>, Å trumbelj, E. and Kononenko, I. (2014) <doi:10.1007/s10115-013-0679-x>. itsdm also provides functions to diagnose variable response, analyze variable importance, draw spatial dependence of variables and examine variable contribution. As utilities, the package includes a few functions to download bioclimatic variables including WorldClim version 2.0 (see Fick, S.E. and Hijmans, R.J. (2017) <doi:10.1002/joc.5086>) and CMCC-BioClimInd (see Noce, S., Caporaso, L. and Santini, M. (2020) <doi:10.1038/s41597-020-00726-5>.

Background - Traditional gene set enrichment analyses are typically limited to a few ontologies and do not account for the interdependence of gene sets or terms, resulting in overcorrected p-values. To address these challenges, we introduce mulea, an R package offering comprehensive overrepresentation and functional enrichment analysis. Results - mulea employs a progressive empirical false discovery rate (eFDR) method, specifically designed for interconnected biological data, to accurately identify significant terms within diverse ontologies. mulea expands beyond traditional tools by incorporating a wide range of ontologies, encompassing Gene Ontology, pathways, regulatory elements, genomic locations, and protein domains. This flexibility enables researchers to tailor enrichment analysis to their specific questions, such as identifying enriched transcriptional regulators in gene expression data or overrepresented protein domains in protein sets. To facilitate seamless analysis, mulea provides gene sets (in standardised GMT format) for 27 model organisms, covering 22 ontology types from 16 databases and various identifiers resulting in almost 900 files. Additionally, the muleaData ExperimentData Bioconductor package simplifies access to these pre-defined ontologies. Finally, mulea's architecture allows for easy integration of user-defined ontologies, or GMT files from external sources (e.g., MSigDB or Enrichr), expanding its applicability across diverse research areas. Conclusions - mulea is distributed as a CRAN R package. It offers researchers a powerful and flexible toolkit for functional enrichment analysis, addressing limitations of traditional tools with its progressive eFDR and by supporting a variety of ontologies. Overall, mulea fosters the exploration of diverse biological questions across various model organisms.

Stochastic block model used for dynamic graphs represented by Poisson processes. To model recurrent interaction events in continuous time, an extension of the stochastic block model is proposed where every individual belongs to a latent group and interactions between two individuals follow a conditional inhomogeneous Poisson process with intensity driven by the individualsâ latent groups. The model is shown to be identifiable and its estimation is based on a semiparametric variational expectation-maximization algorithm. Two versions of the method are developed, using either a nonparametric histogram approach (with an adaptive choice of the partition size) or kernel intensity estimators. The number of latent groups can be selected by an integrated classification likelihood criterion. Y. Baraud and L. Birgé (2009). <doi:10.1007/s00440-007-0126-6>. C. Biernacki, G. Celeux and G. Govaert (2000). <doi:10.1109/34.865189>. M. Corneli, P. Latouche and F. Rossi (2016). <doi:10.1016/j.neucom.2016.02.031>. J.-J. Daudin, F. Picard and S. Robin (2008). <doi:10.1007/s11222-007-9046-7>. A. P. Dempster, N. M. Laird and D. B. Rubin (1977). <http://www.jstor.org/stable/2984875>. G. Grégoire (1993). <http://www.jstor.org/stable/4616289>. L. Hubert and P. Arabie (1985). <doi:10.1007/BF01908075>. M. Jordan, Z. Ghahramani, T. Jaakkola and L. Saul (1999). <doi:10.1023/A:1007665907178>. C. Matias, T. Rebafka and F. Villers (2018). <doi:10.1093/biomet/asy016>. C. Matias and S. Robin (2014). <doi:10.1051/proc/201447004>. H. Ramlau-Hansen (1983). <doi:10.1214/aos/1176346152>. P. Reynaud-Bouret (2006). <doi:10.3150/bj/1155735930>.

Estimation, based on conditional maximum likelihood, of the quadratic exponential model proposed by Bartolucci, F. & Nigro, V. (2010, Econometrica) <DOI:10.3982/ECTA7531> and of a simplified and a modified version of this model. The quadratic exponential model is suitable for the analysis of binary longitudinal data when state dependence (further to the effect of the covariates and a time-fixed individual intercept) has to be taken into account. Therefore, this is an alternative to the dynamic logit model having the advantage of easily allowing conditional inference in order to eliminate the individual intercepts and then getting consistent estimates of the parameters of main interest (for the covariates and the lagged response). The simplified version of this model does not distinguish, as the original model does, between the last time occasion and the previous occasions. The modified version formulates in a different way the interaction terms and it may be used to test in a easy way state dependence as shown in Bartolucci, F., Nigro, V. & Pigini, C. (2018, Econometric Reviews) <DOI:10.1080/07474938.2015.1060039>. The package also includes estimation of the dynamic logit model by a pseudo conditional estimator based on the quadratic exponential model, as proposed by Bartolucci, F. & Nigro, V. (2012, Journal of Econometrics) <DOI:10.1016/j.jeconom.2012.03.004>. For large time dimensions of the panel, the computation of the proposed models involves a recursive function from Krailo M. D., & Pike M. C. (1984, Journal of the Royal Statistical Society. Series C (Applied Statistics)) and Bartolucci F., Valentini, F. & Pigini C. (2021, Computational Economics <DOI:10.1007/s10614-021-10218-2>.

This package provides methods that use flexible variants of multidimensional scaling (MDS) which incorporate parametric nonlinear distance transformations and trade-off the goodness-of-fit fit with structure considerations to find optimal hyperparameters, also known as structure optimized proximity scaling (STOPS) (Rusch, Mair & Hornik, 2023,<doi:10.1007/s11222-022-10197-w>). The package contains various functions, wrappers, methods and classes for fitting, plotting and displaying different 1-way MDS models with ratio, interval, ordinal optimal scaling in a STOPS framework. These cover essentially the functionality of the package smacofx, including Torgerson (classical) scaling with power transformations of dissimilarities, SMACOF MDS with powers of dissimilarities, Sammon mapping with powers of dissimilarities, elastic scaling with powers of dissimilarities, spherical SMACOF with powers of dissimilarities, (ALSCAL) s-stress MDS with powers of dissimilarities, r-stress MDS, MDS with powers of dissimilarities and configuration distances, elastic scaling powers of dissimilarities and configuration distances, Sammon mapping powers of dissimilarities and configuration distances, power stress MDS (POST-MDS), approximate power stress, Box-Cox MDS, local MDS, Isomap, curvilinear component analysis (CLCA), curvilinear distance analysis (CLDA) and sparsified (power) multidimensional scaling and (power) multidimensional distance analysis (experimental models from smacofx influenced by CLCA). All of these models can also be fit by optimizing over hyperparameters based on goodness-of-fit fit only (i.e., no structure considerations). The package further contains functions for optimization, specifically the adaptive Luus-Jaakola algorithm and a wrapper for Bayesian optimization with treed Gaussian process with jumps to linear models, and functions for various c-structuredness indices. Hyperparameter optimization can be done with a number of techniques but we recommend either Bayesian optimization or particle swarm. For using "Kriging", users need to install a version of the archived DiceOptim R package.

Statistical power and minimum required sample size calculations for (1) testing a proportion (one-sample) against a constant, (2) testing a mean (one-sample) against a constant, (3) testing difference between two proportions (independent samples), (4) testing difference between two means or groups (parametric and non-parametric tests for independent and paired samples), (5) testing a correlation (one-sample) against a constant, (6) testing difference between two correlations (independent samples), (7) testing a single coefficient in multiple linear regression, logistic regression, and Poisson regression (with standardized or unstandardized coefficients, with no covariates or covariate adjusted), (8) testing an indirect effect (with standardized or unstandardized coefficients, with no covariates or covariate adjusted) in the mediation analysis (Sobel, Joint, and Monte Carlo tests), (9) testing an R-squared against zero in linear regression, (10) testing an R-squared difference against zero in hierarchical regression, (11) testing an eta-squared or f-squared (for main and interaction effects) against zero in analysis of variance (could be one-way, two-way, and three-way), (12) testing an eta-squared or f-squared (for main and interaction effects) against zero in analysis of covariance (could be one-way, two-way, and three-way), (13) testing an eta-squared or f-squared (for between, within, and interaction effects) against zero in one-way repeated measures analysis of variance (with non-sphericity correction and repeated measures correlation), and (14) testing goodness-of-fit or independence for contingency tables. Alternative hypothesis can be formulated as "not equal", "less", "greater", "non-inferior", "superior", or "equivalent" in (1), (2), (3), and (4); as "not equal", "less", or "greater" in (5), (6), (7) and (8); but always as "greater" in (9), (10), (11), (12), (13), and (14). Reference: Bulus and Polat (2023) <https://osf.io/ua5fc>.

This package provides tools for semantic segmentation of geospatial data using convolutional neural network-based deep learning. Utility functions allow for creating masks, image chips, data frames listing image chips in a directory, and DataSets for use within DataLoaders. Additional functions are provided to serve as checks during the data preparation and training process. A UNet architecture can be defined with 4 blocks in the encoder, a bottleneck block, and 4 blocks in the decoder. The UNet can accept a variable number of input channels, and the user can define the number of feature maps produced in each encoder and decoder block and the bottleneck. Users can also choose to (1) replace all rectified linear unit (ReLU) activation functions with leaky ReLU or swish, (2) implement attention gates along the skip connections, (3) implement squeeze and excitation modules within the encoder blocks, (4) add residual connections within all blocks, (5) replace the bottleneck with a modified atrous spatial pyramid pooling (ASPP) module, and/or (6) implement deep supervision using predictions generated at each stage in the decoder. A unified focal loss framework is implemented after Yeung et al. (2022) <doi:10.1016/j.compmedimag.2021.102026>. We have also implemented assessment metrics using the luz package including F1-score, recall, and precision. Trained models can be used to predict to spatial data without the need to generate chips from larger spatial extents. Functions are available for performing accuracy assessment. The package relies on torch for implementing deep learning, which does not require the installation of a Python environment. Raster geospatial data are handled with terra'. Models can be trained using a Compute Unified Device Architecture (CUDA)-enabled graphics processing unit (GPU); however, multi-GPU training is not supported by torch in R'.

This package provides a variety of methods are provided to estimate and visualize distributional differences in terms of effect sizes. Particular emphasis is upon evaluating differences between two or more distributions across the entire scale, rather than at a single point (e.g., differences in means). For example, Probability-Probability (PP) plots display the difference between two or more distributions, matched by their empirical CDFs (see Ho and Reardon, 2012; <doi:10.3102/1076998611411918>), allowing for examinations of where on the scale distributional differences are largest or smallest. The area under the PP curve (AUC) is an effect-size metric, corresponding to the probability that a randomly selected observation from the x-axis distribution will have a higher value than a randomly selected observation from the y-axis distribution. Binned effect size plots are also available, in which the distributions are split into bins (set by the user) and separate effect sizes (Cohen's d) are produced for each bin - again providing a means to evaluate the consistency (or lack thereof) of the difference between two or more distributions at different points on the scale. Evaluation of empirical CDFs is also provided, with built-in arguments for providing annotations to help evaluate distributional differences at specific points (e.g., semi-transparent shading). All function take a consistent argument structure. Calculation of specific effect sizes is also possible. The following effect sizes are estimable: (a) Cohen's d, (b) Hedges g, (c) percentage above a cut, (d) transformed (normalized) percentage above a cut, (e) area under the PP curve, and (f) the V statistic (see Ho, 2009; <doi:10.3102/1076998609332755>), which essentially transforms the area under the curve to standard deviation units. By default, effect sizes are calculated for all possible pairwise comparisons, but a reference group (distribution) can be specified.

Pancreatic ductal adenocarcinoma (PDA) has a relatively poor prognosis and is one of the most lethal cancers. Molecular classification of gene expression profiles holds the potential to identify meaningful subtypes which can inform therapeutic strategy in the clinical setting. The Pancreatic Cancer Adenocarcinoma Tool-Kit (PDATK) provides an S4 class-based interface for performing unsupervised subtype discovery, cross-cohort meta-clustering, gene-expression-based classification, and subsequent survival analysis to identify prognostically useful subtypes in pancreatic cancer and beyond. Two novel methods, Consensus Subtypes in Pancreatic Cancer (CSPC) and Pancreatic Cancer Overall Survival Predictor (PCOSP) are included for consensus-based meta-clustering and overall-survival prediction, respectively. Additionally, four published subtype classifiers and three published prognostic gene signatures are included to allow users to easily recreate published results, apply existing classifiers to new data, and benchmark the relative performance of new methods. The use of existing Bioconductor classes as input to all PDATK classes and methods enables integration with existing Bioconductor datasets, including the 21 pancreatic cancer patient cohorts available in the MetaGxPancreas data package. PDATK has been used to replicate results from Sandhu et al (2019) [https://doi.org/10.1200/cci.18.00102] and an additional paper is in the works using CSPC to validate subtypes from the included published classifiers, both of which use the data available in MetaGxPancreas. The inclusion of subtype centroids and prognostic gene signatures from these and other publications will enable researchers and clinicians to classify novel patient gene expression data, allowing the direct clinical application of the classifiers included in PDATK. Overall, PDATK provides a rich set of tools to identify and validate useful prognostic and molecular subtypes based on gene-expression data, benchmark new classifiers against existing ones, and apply discovered classifiers on novel patient data to inform clinical decision making.