4.0 Article

Model-based random forests for ordinal regression

Journal

INTERNATIONAL JOURNAL OF BIOSTATISTICS
Volume 16, Issue 2, Pages -

Publisher

WALTER DE GRUYTER GMBH
DOI: 10.1515/ijb-2019-0063

Keywords

amyotrophic lateral sclerosis; conditional distribution function; conditional odds function; ordinal outcome; random forest; transformation model

Funding

  1. Horizon 2020 Research and Innovation Programme of the European Union [681094]
  2. Swiss State Secretariat for Education, Research and Innovation (SERI) [15.0137]
  3. Swiss National Science Foundation [200021_184603]
  4. Swiss National Science Foundation (SNF) [200021_184603] Funding Source: Swiss National Science Foundation (SNF)

Ask authors/readers for more resources

We study and compare several variants of random forests tailored to prognostic models for ordinal outcomes. Models of the conditional odds function are employed to understand the various random forest flavours. Existing random forest variants for ordinal outcomes, such as Ordinal Forests and Conditional Inference Forests, are evaluated in the presence of a non-proportional odds impact of prognostic variables. We propose two novel random forest variants in the model-based transformation forest family, only one of which explicitly assumes proportional odds. These two novel transformation forests differ in the specification of the split procedures for the underlying ordinal trees. One of these split criteria is able to detect changes in non-proportional odds situations and the other one focuses on finding proportional-odds signals. We empirically evaluate the performance of the existing and proposed methods using a simulation study and illustrate the practical aspects of the procedures by a re-analysis of the respiratory sub-item in functional rating scales of patients suffering from Amyotrophic Lateral Sclerosis (ALS).

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.0
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

Editorial Material Biodiversity Conservation

Insights from regional and short-term biodiversity monitoring datasets are valuable: a reply to Daskalova et al. 2021

Sebastian Seibold, Torsten Hothorn, Martin M. Gossner, Nadja K. Simons, Nico Bluthgen, Jorg Muller, Didem Ambarli, Christian Ammer, Jurgen Bauhus, Markus Fischer, Jan C. Habel, Caterina Penone, Peter Schall, Ernst-Detlef Schulze, Wolfgang W. Weisser

Summary: Existing datasets on insect biodiversity are often limited, and while new monitoring programs have been initiated, they still cover rather short time periods. Some researchers argue that temporal trends of insect populations derived from short time series are biased towards extreme trends, while others suggest that there is no overall decline in insect population.

INSECT CONSERVATION AND DIVERSITY (2021)

Article Statistics & Probability

Multivariate conditional transformation models

Nadja Klein, Torsten Hothorn, Luisa Barbanti, Thomas Kneib

Summary: Regression models are an important aspect of contemporary regression analysis, but often rely on simplistic assumptions. We propose a general framework for multivariate conditional transformation models that can describe the entire distribution in a tractable and interpretable yet flexible way, allowing for non-linear effects of covariates and likelihood-based inference.

SCANDINAVIAN JOURNAL OF STATISTICS (2022)

Article Critical Care Medicine

Unbiased Recursive Partitioning Enables Robust and Reliable Outcome Prediction in Acute Spinal Cord Injury

Muriel Buri, Lorenzo G. Tanadini, Torsten Hothorn, Armin Curt

Summary: URP-CTREE, as a clinical guidance tool, demonstrates high accuracy in predicting future clinical outcomes, showing robustness and replicability in identifying homogeneous subgroups.

JOURNAL OF NEUROTRAUMA (2022)

Article Computer Science, Artificial Intelligence

Deep and interpretable regression models for ordinal outcomes

Lucas Kook, Lisa Herzog, Torsten Hothorn, Oliver Durr, Beate Sick

Summary: This study introduces ordinal neural network transformation models that combine deep learning with classical ordinal regression approaches, achieving a balance between flexibility and interpretability for image and tabular data modeling, with faster training speed than traditional DL models.

PATTERN RECOGNITION (2022)

Letter Hematology

Crovalimab for treatment of patients with paroxysmal nocturnal haemoglobinuria and complement C5 polymorphism: Subanalysis of the phase 1/2 COMPOSER study

Jun-ichi Nishimura, Kensuke Usuki, Julia Ramos, Satoshi Ichikawa, Muriel Buri, Anna Kiialainen, Alexandre Sostelly, Regis Peffault de Latour, Ido Paz-Priel, Alexander Roth

BRITISH JOURNAL OF HAEMATOLOGY (2022)

Article Ecology

Diverse Effects of Climate, Land Use, and Insects on Dung and Carrion Decomposition

Jana Englmeier, Oliver Mitesser, M. Eric Benbow, Torsten Hothorn, Christian von Hoermann, Caryl Benjamin, Ute Fricke, Cristina Ganuza, Maria Haensel, Sarah Redlich, Rebekka Riebl, Sandra Rojas Botero, Thomas Rummler, Ingolf Steffan-Dewenter, Elisa Stengel, Cynthia Tobisch, Johannes Uhler, Lars Uphus, Jie Zhang, Jorg Mueller

Summary: Land-use intensification and climate change pose threats to ecosystem functions, especially the decomposition of necromass. This study measured the decomposition of rat carrion and bison dung across a climate gradient and different land-use types, finding that decomposition of dung was influenced by elevation while carrion decomposition was affected by both elevation and temperature. Insects played a significant role in carrion decay but had no effect on dung decomposition. The findings highlight the potential impacts of climate change and insect decline on decomposition processes and ecosystem self-regulation.

ECOSYSTEMS (2023)

Article Mathematical & Computational Biology

On the relevance of prognostic information for clinical trials: A theoretical quantification

Sandra Siegfried, Stephen Senn, Torsten Hothorn

Summary: The question of how to leverage individual patient data for designing more powerful clinical trials becomes increasingly important in the era of digitalization. Incorporating historical information in the design and analysis of future clinical trials can lead to smaller yet equally powerful studies. A study found that adjusting the analysis with respect to a prognostic score obtained from historical data can significantly reduce the required sample size.

BIOMETRICAL JOURNAL (2023)

Article Biodiversity Conservation

A comparison of different Malaise trap types

Johannes Uhler, Peter Haase, Lara Hoffmann, Torsten Hothorn, Jurgen Schmidl, Stefan Stoll, Ellen A. R. Welti, Jorn Buse, Jorg Mueller

Summary: This study compared the effects of different Malaise trap types on insect biomass and species richness. It found that trap type significantly influences the captured insect biomass and species richness. The Townes trap models showed higher total species richness compared to Bartak traps, and the white roofed Townes trap captured a higher richness of pollinators.

INSECT CONSERVATION AND DIVERSITY (2022)

Article Mathematical & Computational Biology

A transformation perspective on marginal and conditional models

Luisa Barbanti, Torsten Hothorn

Summary: This paper presents a novel model for clustered observations, which can be applied to various types of response variables. The model is demonstrated using sleep deprivation and toe nail data, and the analysis of two clinical trials is discussed. Overall, the paper provides a new approach for analyzing clustered observations.

BIOSTATISTICS (2022)

Article Biodiversity Conservation

Enhancing the structural diversity between forest patches-A concept and real-world experiment to study biodiversity, multifunctionality and forest resilience across spatial scales

Joerg Mueller, Oliver Mitesser, Marc W. Cadotte, Fons van der Plas, Akira S. Mori, Christian Ammer, Anne Chao, Michael Scherer-Lorenzen, Petr Baldrian, Claus Baessler, Peter Biedermann, Simone Cesarz, Alice Classen, Benjamin M. Delory, Heike Feldhaar, Andreas Fichtner, Torsten Hothorn, Claudia Kuenzer, Marcell K. Peters, Kerstin Pierick, Thomas Schmitt, Bernhard Schuldt, Dominik Seidel, Diana Six, Ingolf Steffan-Dewenter, Simon Thorn, Goddert von Oheimb, Martin Wegmann, Wolfgang W. Weisser, Nico Eisenhauer

Summary: Intensification of land use by humans has led to a homogenization of landscapes and decreasing resilience of ecosystems globally due to a loss of biodiversity, including the majority of forests. Biodiversity-ecosystem functioning (BEF) research has provided compelling evidence for a positive effect of biodiversity on ecosystem functions and services at the local (alpha-diversity) scale, but we largely lack empirical evidence on how the loss of between-patch beta-diversity affects biodiversity and multifunctionality at the landscape scale (gamma-diversity).

GLOBAL CHANGE BIOLOGY (2023)

Article Computer Science, Theory & Methods

Probabilistic time series forecasts with autoregressive transformation models

David Ruegamer, Philipp F. M. Baumann, Thomas Kneib, Torsten Hothorn

Summary: Probabilistic forecasting of time series is a crucial task in various applications and research domains. This paper introduces Autoregressive Transformation Models (ATMs), which aim to combine expressive distributional forecasts with an interpretable model specification by utilizing a semi-parametric distribution assumption. The properties of ATMs are demonstrated through both theoretical analysis and empirical evaluation on simulated and real-world forecasting datasets.

STATISTICS AND COMPUTING (2023)

Article Statistics & Probability

Distribution-Free Location-Scale Regression

Sandra Siegfried, Lucas Kook, Torsten Hothorn

Summary: We introduce a generalized additive model for location, scale, and shape (GAMLSS) for distribution-free and parsimonious regression modelling. The model replaces strict parametric distribution with a transformation function and limits the number of linear or smooth model terms. Likelihood and score functions are derived for various types of observations. Various algorithms are used for model estimation and parameter interpretability is connected to model selection. A novel best subset selection procedure is proposed for simpler interpretation. Numerous applications are provided as examples and all analyses are reproducible using the tram add-on package to the R system.

AMERICAN STATISTICIAN (2023)

Article Health Care Sciences & Services

Estimating transformationsfor evaluating diagnostic testswith covariate adjustment

Ainesh Sewak, Torsten Hothorn

Summary: Receiver operating characteristic analysis is widely used for evaluating the accuracy of medical diagnostic tests. However, there is no consensus on a single framework for consistent statistical inference in handling complex medical data. We propose a regression model that accommodates these complexities and provides unbiased estimates with nominal coverage.

STATISTICAL METHODS IN MEDICAL RESEARCH (2023)

Article Clinical Neurology

Inclusive Trial Designs in Acute Spinal Cord Injuries: Prediction-Based Stratification of Clinical Walking Outcome and Projected Enrolment Frequencies

Adrian Cathomen, Laura Sirucek, Tim Killeen, Rainer Abel, Doris Maier, Norbert Weidner, Rudiger Rupp, Torsten Hothorn, John D. Steeves, Armin Curt, Marc Bolliger

Summary: This study used URP-CTREE to analyze data from 309 SCI patients, identifying homogeneous outcome cohorts, which were validated by an internal validation group. This method can help design clinical trials with broad inclusion criteria and improve the quality of research design.

NEUROREHABILITATION AND NEURAL REPAIR (2022)

Article Mathematical & Computational Biology

Individual participant data meta-analysis with mixed-effects transformation models

Balint Tamasi, Michael Crowther, Milo Alan Puhan, Ewout W. Steyerberg, Torsten Hothorn

Summary: This study introduces a new model class for meta-analysis of individual participant data (IPD) that incorporates general normally distributed random effects into linear transformation models and can handle data with arbitrary random censoring patterns. The proposed method shows efficiency in predicting the survival of patients with chronic obstructive pulmonary disease, compared to alternative approaches. Simulation study confirms the correctness of the implementation and highlights the efficiency of the proposed method.

BIOSTATISTICS (2022)

No Data Available