TRThomas Rusch

Thomas Rusch

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Thomas Rusch has a total of 28 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • divoRce: Diagnosing Separation Phenomena in Latent Polyhedral Categorical Response Models

    by Thomas Rusch, Lukas Sablica, Kurt Hornik

    Utilities for checking and diagnosing separation phenomena (complete separation, quasi-complete separation, overlap) for latent polyhedral categorical response models including binary, nominal and ordinal data models with link construction as baseline-category, adjacent-category, cumulative, sequential (continuation-ratio), ordered stereotype and arbitrary link error distribution (including logit, probit, cauchit, cloglog, loglog, log amongst others) based on structure vectors as described in Sablica, Hornik, Rusch (2026) and Rusch, Sablica, Hornik (2026) . The package functionality includes a check for separation and a check for overlap with a linear program, a quick check for overlap, a sequential and a parallel check for overlap based on subsamples, detailed separation diagnostics such as identification of columns in the design matrix/structure vector matrix that lead to separation, identification of design matrix/structure vector matrix rows that lead to separation, the dimension of the recession cone and the structure vectors that are linearities. Calculations can be done with package 'rcdd' to employ arbitrary precision and various other linear program solvers can also be used via package 'ROI'. Implemented checks can be done before model fitting based on outcome and design matrix or structure vector matrix or via formula (pre-fit), or after model fitting for objects of class 'glm', 'multinom', 'polr', 'osm', 'clm', 'bracl', 'brmultinom'. Wrappers for other classes can be easily constructed by third parties via pre-fit functions.

  • Open Access

    Multidimensional Scaling With Heaviside Weighting: Extensions to Curvilinear Component Analysis and Curvilinear Distance Analysis

    by Thomas Rusch

    In this article, I propose extensions to multidimensional scaling (MDS) variants that use Heaviside weighting for manifold learning, namely, curvilinear component analysis and curvilinear distance analysis. These methods weight the approximation error with a Heaviside function to ignore larger fitted distances in the configuration, thus effectively providing a localized version of MDS. I extend these methods with (i) optimal scaling that allows for linear, affine or spline transformations of the proximities and (ii) fitting power transformations of proximities and distances. I suggest a quasi‐MM algorithm for optimization and present the methods either for a fixed neighbourhood parameter or as self‐organizing. I employ both geodesic distances and arbitrary dissimilarities as proximities. This yields new MDS variants that I call extended curvilinear component analysis (eCLCA), extended curvilinear power component analysis (eCLPCA), extended curvilinear distance analysis (eCLDA) and extended curvilinear power distance analysis (eCLPDA). With three data examples, I show how the extensions can improve learning a manifold.


  • Testing the Shamiri Intervention and Its Components With Kenyan Adolescents During the COVID-19 Pandemic: Outcomes of a Universal, 5-Arm Randomized Controlled Trial

    by Katherine E. Venturo-Conerly, Tom L. Osborn, Thomas Rusch, Brenda Kemuma Ochuku, Natalie E. Johnson, Afra van der Markt, Christine M. Wasanga, John R. Weisz


  • Open Access

    Testing pathways to scale: study protocol for a three-arm randomized controlled trial of a centralized and a decentralized (“Train the Trainers”) dissemination of a mental health program for Kenyan adolescents

    by Brenda Ochuku, Tom L. Osborn, Daisy Nerima, Afra van der Markt, Thomas Rusch, Herman Omune, Solace Akello, David M. Ndetei, Katherine E. Venturo-Conerly

    Background: Providing care in Kenya to all youth in need is difficult because of a shortage of professional providers and societal stigma. Previous trials of the Anansi model, which involves delivering low-touch mental health interventions through a tiered caregiving model (including lay-providers, supervisors, and clinical experts), have shown its effectiveness for reducing depression and anxiety symptoms in school-going Kenyan adolescents. In this trial, we aim to assess two different scale-up strategies by comparing centralized implementation (i.e., by the organization that designed the Anansi model) against implementation through an implementing partner. Methods: In this three-arm trial, 1600 adolescents aged 13 to 20 years will be randomized to receive the Shamiri intervention from either the Shamiri Institute or an implementation partner or to be placed in the treatment as usual (TAU) control group. The implementation partner will be trained and supplied with protocols to ensure that the same procedures are followed by both implementors. Implementation activities will run concurrently for both implementors. The Shamiri intervention will be delivered by trained lay providers to groups of 10–15 adolescents over four weekly sessions which will take place in secondary schools in Machakos and Makueni counties in Kenya. The TAU group will receive the usual care offered by their respective schools. Outcomes will be assessed at baseline, midpoint (2 weeks), endpoint (4 weeks), and 1 month follow-up. The analysis will be based on an intent-to-treat approach. Mixed effects models will be used to assess trajectories over time of the primary outcomes (anxiety and depressive symptoms, mental well-being, perceived social support, and academic performance) and secondary outcomes for the intervention groups and the control group. Effect sizes will be computed for the mean differences of the intervention and control arms at midpoint, endpoint, and follow-up. Discussion: This trial will provide insight into the comparative effectiveness of different strategies for scaling a school-based mental health care model. Findings will also indicate areas for improved efficiency of the model to enhance its replicability by other implementors. Trial registration: Pan African Clinical Trials Registry (PACTR) (ID: PACTR202305589854478, Approved: 02/05/2023).


  • Open Access

    COPS in Action: Exploring Structure in the Usage of the Youth Psychotherapy MATCH

    by Thomas Rusch, Katherine Venturo-Conerly, Gioia Baja, Patrick Mair

    This article is an introduction to Cluster Optimized Proximity Scaling (COPS) aimed at practitioners, as well as a tutorial on the usage of the corresponding R package cops. COPS is a variant of multidimensional scaling (MDS) that aims at providing a clustered configuration while still representing multivariate dissimilarities faithfully. It subsumes most popular MDS versions as special cases. We illustrate the ideas, use, flexibility and versatility of the method and the package with data from clinical psychology on how modules of the Modular Approach to Therapy for Children (MATCH) are used by clinicians in the wild. We supplement the COPS analyses with density-based hierarchical clustering in the original space and faceting with support vector machines. We find that scaling with COPS gives a sensible and insightful spatial arrangement of the modules, allows easy identification of clusters of modules and provides clear facets of modules corresponding to the MATCH protocols. In that respect COPS works better than both standard MDS and clustering.


  • Open Access

    Structure-based hyperparameter selection with Bayesian optimization in multidimensional scaling

    by Thomas Rusch, Patrick Mair, Kurt Hornik

    We introduce the structure optimized proximity scaling (STOPS) framework for hyperparameter selection in parametrized multidimensional scaling and extensions (proximity scaling; PS). The selection process for hyperparameters is based on the idea that we want the configuration to show a certain structural quality (c-structuredness). A number of structures and how to measure them are discussed. We combine the structural quality by means of c-structuredness indices with the PS badness-of-fit measure in a multi-objective scalarization approach, yielding the Stoploss objective. Computationally we suggest a profile-type algorithm that first solves the PS problem and then uses Stoploss in an outer step to optimize over the hyperparameters. Bayesian optimization with treed Gaussian processes as a an apt and efficient strategy for carrying out the outer optimization is recommended. This way, hyperparameter tuning for many instances of PS is covered in a single conceptual framework. We illustrate the use of the STOPS framework with three data examples.


  • Open Access

    Cluster Optimized Proximity Scaling

    by Thomas Rusch, Patrick Mair, Kurt Hornik

    Proximity scaling methods such as Multidimensional Scaling (MDS) represent objects in a low dimensional configuration so that fitted object distances optimally approximate object proximities. Besides finding the optimal configuration, an additional goal may be to make statements about the cluster arrangement of objects. This fails if the configuration lacks appreciable clusteredness. We present Cluster Optimized Proximity Scaling (COPS), which attempts to find a configuration that exhibits clusteredness. In COPS, a flexible parametrized scaling loss function that may emphasize differentiation information in the proximities is augmented with an index (OPTICS Cordillera) that penalizes lack of clusteredness of the configuration. We present two variants of this, one for finding a configuration directly and one for hyperparameter selection for parametric stresses. We apply both to a functional magnetic resonance imaging (fMRI) data set on neural representations of mental states in a social cognition task and show that COPS improves clusteredness of the configuration, enabling visual identification of clusters of mental states. Online supplementary material is available including an R package and a document with additional details.