Všechny publikace

Differential Item Functioning (DIF) analysis is used to identify potentially biased items in multi-item measurements. In addition to testing the statistical significance, it is essential to evaluate t…

Differential Item Functioning (DIF) analysis is used to identify potentially biased items in multi-item measurements. In addition to testing the statistical significance, it is essential to evaluate the practical significance of DIF through effect size measures. We review existing DIF effect size measures and cut-off values used to classify the effect size magnitudes for the Mantel-Haenszel test, SIBTEST, and logistic regression for binary items, and introduce a refinement of area-based effect size measures. A simulation study is conducted to investigate the properties of these effect size measures and existing classification guidelines, and to assess their comparative performance. The results indicate that some commonly used effect size measures exhibit undesirable properties, including inconsistent classifications, systematic underestimation of the magnitude of the underlying DIF, and strong dependence on design factors. To address these issues, we introduce usage restrictions for some effect size measures, revised cut-off values that unify results across different methods, and propose new cut-off values for area-based effect size measures. The methods are demonstrated using two real data examples. Implementation is provided in the R software.

This repository contains the dataset and accompanying code used in the manuscript: "Nonlinear Dynamics in Intensive Longitudinal Health Behavior Data: Methodological Rethinking using the 4HAIE Data." …

This repository contains the dataset and accompanying code used in the manuscript: "Nonlinear Dynamics in Intensive Longitudinal Health Behavior Data: Methodological Rethinking using the 4HAIE Data." The dataset originates from the 4HAIE (Healthy Aging in Industrial Environment - Program 4) study, a 12-month intensive longitudinal study designed to examine within-person dynamics of health behaviors and well-being in everyday life (see corresponding protocol paper Elavsky et al., 2021). Dataset Contents The repository includes: Merged analysis dataset linking daily physical activity with corresponding end-of-day life satisfaction assessments. Daily physical activity data collected continuously over approximately 12 months using Fitbit Charge 3 and Fitbit Charge 4 devices. Physical activity is represented by daily step counts. Ecological Momentary Assessment (EMA) data collected during four 14-day measurement bursts distributed across the 12-month study period. Participants completed an end-of-day survey assessing life satisfaction using two items adapted from the Satisfaction with Life Scale (SWLS). Illustrative analysis code demonstrating the generalized additive modeling (GAM) approach used to investigate nonlinear relationships between physical activity and life satisfaction over time. Study Design Participants were monitored continuously for physical activity throughout the study while completing repeated intensive EMA assessments during four separate two-week measurement bursts. This design enabled examination of both long-term behavioral patterns and short-term fluctuations in subjective well-being within individuals. Variables The repository includes variables related to: Participant identifier (de-identified) Date of observation Daily step count (Fitbit) End-of-day life satisfaction ratings (two EMA items) Additional temporal variables required for longitudinal modeling Variable definitions and coding information are provided in the accompanying documentation. Purpose The dataset is intended to illustrate statistical approaches for analyzing intensive longitudinal health behavior data, with particular emphasis on nonlinear modeling of within-person associations using generalized additive models (GAMs). Rather than serving solely as an empirical example, the dataset demonstrates a methodological framework for capturing complex temporal relationships that may not be adequately represented using traditional linear approaches. Associated Publication This dataset accompanies the manuscript: Nonlinear Dynamics in Intensive Longitudinal Health Behavior Data: Methodological Rethinking using the 4HAIE Data. If you use this dataset or the accompanying analysis code, please cite both this Zenodo repository and the associated publication.   References: ELavsky, S., Brabec, M., Maly, M. (in review). Nonlinear Dynamics in Intensive Longitudinal Health Behavior Data: Methodological Rethinking using the 4HAIE Data. Elavsky S, Jandačková V, Knapová L, Vašendová V, Sebera M, Kaštovská B, Blaschová D, Kühnová J, Cimler R, Vilímek D, Bosek T, Koenig J, Jandačka D. Physical activity in an air-polluted environment: behavioral, psychological and neuroimaging protocol for a prospective cohort study (Healthy Aging in Industrial Environment study - Program 4). BMC Public Health. 2021 Jan 12;21(1):126. doi: 10.1186/s12889-021-10166-4. PMID: 33435943; PMCID: PMC7801866. Elavsky S, Brabec M, Maly M, Knapova L, Kastovska B, Sebera M, Ely M, Jandackova VK, Keller J, Pavel M. The temporal dynamics of the association between daily physical activity and life satisfaction. Ann Behav Med. 2025 Jan 4;59(1):kaaf079. doi: 10.1093/abm/kaaf079. PMID: 41224246; PMCID: PMC12757008.

One of the main challenges when developing medical decision support systems for the emergency room is adequately filtering the most relevant information. High workload, stress, and the necessity for u…

One of the main challenges when developing medical decision support systems for the emergency room is adequately filtering the most relevant information. High workload, stress, and the necessity for urgent decisions require precise answers to the questions posed. Although LLM-based systems can provide abundant information, physicians need concise and relevant data in this particular clinical setting. In this study, we perform a pilot assessment of the transparency of selected LLM-based systems. The comparative analysis includes ChatGPT o1 model, which was asked to produce responses with varying temperatures and a pilot graph-based RAG specializing in cardiovascular diseases. A survey was conducted among 33 clinicians regarding the amount of information contained in the provided prompts. Physicians favored the most readable, specific, and helpful answers in emergency department conditions. Reliable medical data and the form in which answers are delivered are crucial for physicians working in the emergency room. We conclude that physicians have preferences for LLM responses at a specific temperature. Further research should be expanded to enable tailoring responses not only to the clinical situation but also to the experience of the asking physician.

This study explores the use of explainable artificial intelligence in education, with a focus on its relevance for Learning Analytics. The research introduces a prototype-based dynamic incremental cla…

This study explores the use of explainable artificial intelligence in education, with a focus on its relevance for Learning Analytics. The research introduces a prototype-based dynamic incremental classification algorithm, Dynamic Incremental Semi-Supervised Fuzzy C-Means (DISSFCM), which leverages fuzzy logic to analyze student interaction data from virtual learning platforms, even when the data are only partially labeled. The proposed methodology generates human-centered explanations by extracting IF-THEN fuzzy rules from the evolving prototypes produced by DISSFCM over successive time intervals. These explanations, expressed in linguistic terms, remain accessible to non-expert stakeholders and are particularly suitable for educational contexts. The Open University Learning Analytics Dataset (OULAD) is utilized for experimentation and validation, providing a realistic scenario for semi-supervised data collection. Visual summaries of the evolving fuzzy rules support the identification of temporal patterns in streaming data. Results show that the model effectively adapts to concept drift while maintaining interpretability. Most notably, it proves robust in handling partially labeled data and variable time granularities, two challenges frequently encountered in real-world Learning Analytics sce- narios. The ability to both predict student outcomes and provide intelligible explanations under such constraints highlights the practical value of the approach. To evaluate the quality and relevance of the generated explanations, an expert-based evaluation was conducted. Domain experts evaluated the clarity, usefulness, and accuracy of the explanations in terms of their support for human understanding and decision-making. The results suggest that the explanations were perceived as generally informative and useful, supporting the method’s relevance for human-centered educational applications. 

We apply an evolving granular-computing modeling approach, called evolving Optimal Granular System (eOGS), to bipolar mood disorder (BD) diagnosis based on speech data streams. The eOGS online learnin…

We apply an evolving granular-computing modeling approach, called evolving Optimal Granular System (eOGS), to bipolar mood disorder (BD) diagnosis based on speech data streams. The eOGS online learning algorithm reveals information granules in the flow and design the structure and parameters of a granular rule-based model with a certain degree of interpretability based on acoustic attributes obtained from phone calls made over 7 months to the Psychiatry department of a hospital. A multi-objective programming problem that trades-off information specificity, model compactness, and numerical and granular error indices is presented. Spectral and prosodic attributes are ranked and selected based on a hybrid Pearson-Spearman correlation coefficient. Low attribute-class correlation, ranging from 0.03 to 0.07, is observed, as well as high class overlap, which is typical in the psychiatric field. eOGS models for BD recognition overcome alternative computational-intelligence models, namely, Dynamic Evolving Neural-Fuzzy Inference System (DENFIS) and Fuzzy-set-Based evolving Modeling (FBeM-Gauss), by a small margin in both best and average cases; followed by eXtended Takagi-Sugeno (xTS) and evolving Takagi Sugeno (eTS) types of models. The proposed eOGS model using only 8 of the original acoustic attributes, and about 15 ‘If-Then’ inference rules, has exhibited the best root mean square error, 0.1361, and 91.8% accuracy in sharp BD class estimates. Granules associated to linguistic labels and a granular input-output map offer human understandability with relation to the inherent process of generating class estimates. Linguistically readable eOGS rules may assist physicians in explaining symptoms and making a diagnosis. 

Systems of partial fuzzy relational equations employing undefined values in the antecedents and consequents have been approached recently. The primary focus was on the issues of sufficient solvability…

Systems of partial fuzzy relational equations employing undefined values in the antecedents and consequents have been approached recently. The primary focus was on the issues of sufficient solvability and solvability criteria. This study introduces another perspective, investigating the behavior of solvability degrees of these systems. We employ operations from the Lower estimation and Dragonfly partial algebras developed in the partial fuzzy set theory framework. Initially, we establish a degree of solvability in an appropriate space of approximations containing potential solutions for the systems. Subsequently, we introduce the concept of the alpha-lift for a given partial fuzzy set and provide its fundamental properties. This concept is employed to modify the antecedents and consequents of a given system of partial fuzzy relational equations, resulting in a modified system. The solvability degree of this modified system is then studied, and we demonstrate that, under sufficient conditions, it significantly enhances the solvability degree of the initial system. This positive impact is observed in the Godel algebra, the underlying algebraic structure of partial algebras. In conclusion, we provide illustrative examples that effectively demonstrate the theoretical results.

nuggets is a framework for subgroup discovery, contrast and emerging patterns, association rules, and more. Developed as a package for the R statistical environment, nuggets provides a novel and …

nuggets is a framework for subgroup discovery, contrast and emerging patterns, association rules, and more. Developed as a package for the R statistical environment, nuggets provides a novel and extensible toolkit for performing rule-based analyses. Both crisp (Boolean) and fuzzy data are supported. The package generates conditions in the form of elementary conjunctions, evaluates them on a dataset, and checks the induced sub-data for interesting statistical properties. A user defined function may be evaluated on generated sub-dataset, which provides a novel generality. The aim of this paper is to present that free software to the soft computing community, as the tool could be useful to both researchers and analysts in the domain of pattern mining, as, besides searching for various existing pattern types, brand new ideas may be easily implemented and evaluated within that framework.

It is already more than 100 years since the first proposal on three-valued logic appeared and it became a seminal work initiating lots of followers among scholars and researchers. Since then, we have …

It is already more than 100 years since the first proposal on three-valued logic appeared and it became a seminal work initiating lots of followers among scholars and researchers. Since then, we have observed distinct logical and algebraic approaches to modeling undefined values, i.e., the situation when the truth value of a given proposition is neither true nor false, but it is not defined. These various algebraic models of three-valued functionality are built to model various types of undefinedness, e.g., conceptional undefinedness, inconsistencies, indeterminable values, meaningless values, or half-true. It is not surprising that recently, these three-valued logics have been extended to partial fuzzy logics, i.e. specific many-valued logics that are extended by the dummy value * that models the undefined truth value. The algebraic structures for such logics are called partial algebras. Recently, two partial algebras, namely the Dragonfly algebra and the Lower Estimation, were both developed to capture the missing or unknown values. Their main idea consists in determining the lower boundary of the truth value of a proposition that we may guarantee after processing the operations independently on what values would replace the dummies *. Such an approach naturally leads to the consequence that the dummy * behaves as a “nearly zero” or “almost false” value. Though the application potential of such algebras in processing the missing values turned out to be very useful at some problems, it turned to be promising to consider a nearly dual approach. Such an approach should model the upper boundary idea and lead to a “nearly one” or “almost true” value. This study provides the first definition of such an algebra and investigates which of the standard properties of residuated lattices remain preserved. Unlike in the lower boundary case, we surprisingly show that in principle all of them are preserved, i.e., that the Upper Boundary algebra, though extended, remains to be the residuated lattice.