Structural Equation Modeling (SEM)

Structural equation modeling (SEM) is a powerful statistical technique to analyze complex relationships between observed and latent variables in psychological research. It allows researchers to test theoretical models about how different psychological constructs relate to one another, while accounting for measurement error and multiple simultaneous pathways of influence.

In psychology, SEM is particularly valuable for studying phenomena that cannot be directly observed, such as intelligence, depression, or motivation, by modeling these constructs through multiple observed indicators (like test scores or questionnaire responses). The technique is widely used in fields like personality psychology to understand how traits relate to behaviors, in developmental psychology to track changes over time, and in social psychology to model complex interpersonal processes.

Our program specializes in advancing SEM methodology through:

• Development of novel estimation methods for complex longitudinal data
• Integration of machine learning techniques with traditional SEM approaches
• Creation of tools for handling missing data and non-normal distributions
• Application of SEM to big data and high-dimensional problems
• Improving psychological measurement for inter and intraindividual analysis

Faculty: Steven Boker, M. Joseph Meyer, Cynthia Tong

Network Science

Graph theory in psychology provides a mathematical framework for analyzing and visualizing relationships between different psychological elements, whether they're symptoms, behaviors, neural connections, or social interactions. Our program leads innovative research in:

• Development of new network psychometric methods, such as Exploratory Graph Analysis, Dynamic Exploratory Graph Analysis, and related techniques
• Application of network analysis to psychological data
• Integration of temporal dynamics in psychological networks
• Advanced visualization techniques for complex psychological data
• New estimation methods
• Software Development

Our faculty pioneered several network-based approaches that have become standard tools in psychological research, particularly in the study of psychopathology and social relationships.

Faculty: Hudson Golino, Teague Rhine Henry

Psychometrics

Psychometrics is the field of study concerned with the theory and methodology of psychological measurement, which includes the design, development, and validation of measurement instruments and the analysis of measurement data. This discipline encompasses both the theoretical frameworks for how to measure psychological attributes (like intelligence, personality traits, or clinical symptoms) and the practical tools for doing so reliably and validly. Psychometricians develop and evaluate psychological tests, questionnaires, and other assessment tools using sophisticated statistical methods such as factor analysis, item response theory, and classical test theory. Their work is crucial across many areas of psychology - from educational testing (where psychometric principles guide the development of standardized tests) to clinical assessment (where accurate measurement of psychological symptoms is vital for diagnosis and treatment planning) to organizational psychology (where psychometric tools help in employee selection and performance evaluation). The field also addresses fundamental questions about the nature of psychological measurement itself, such as how to quantify abstract constructs, how to ensure measurements are consistent across different populations and contexts, and how to determine whether a measurement tool is actually measuring what it claims to measure.

Our psychometrics research focuses on advancing both theoretical frameworks and practical applications for psychological measurement. Key areas include:

• Modern test theory and item response modeling
• Computerized adaptive testing
• Network psychometric approaches to test construction
• Integration of machine learning with traditional psychometric methods
• Development of novel reliability and validity assessment techniques

We maintain strong connections with testing organizations and actively contribute to advancing measurement practices in educational, clinical, and organizational settings.

Faculty: Steven Boker, M. Joseph Meyer, Hudson Golino, Teague Rhine Henry, Karen Schmidt

Human Brain Mapping

Brain imaging data analytics involves sophisticated computational methods to process, analyze, and interpret data from neuroimaging techniques like fMRI, EEG, and PET scans. This field combines elements of signal processing, statistical analysis, machine learning, and network theory to transform raw neuroimaging data into meaningful insights about brain structure and function. In psychology, these analytics enable researchers to study neural correlates of cognitive processes, emotional states, and behavioral patterns. For instance, researchers can examine how different brain regions activate and interact during specific tasks, track changes in brain connectivity associated with learning or development, and identify neural signatures of psychological conditions. Advanced techniques like multivariate pattern analysis (MVPA) and dynamic causal modeling (DCM) allow psychologists to decode mental states from brain activity patterns and understand causal relationships between brain regions. These analytical approaches have revolutionized our understanding of phenomena like decision-making, memory formation, emotional regulation, and various psychological disorders, bridging the gap between observable behavior and its underlying neural mechanisms.

Our program specializes in developing and applying advanced computational methods for analyzing neuroimaging data. Current research focuses on:

• Novel methods for processing and analyzing fMRI, EEG, and PET data
• Development of machine learning approaches for neuroimaging analysis
• Integration of network science with brain mapping
• Methods for analyzing brain connectivity and dynamics
• Tools for large-scale neuroimaging data management and analysis

Faculty: John Darrell Van Horn, Teague Rhine Henry

Non-linear Dynamics

Non-linear dynamics is the study of systems in which relationships between variables are not proportional, leading to complex behaviors such as feedback loops, chaos, and emergent phenomena. In psychology, non-linear dynamics is used to model and analyze processes that evolve over time in unpredictable or irregular ways, such as emotional regulation, cognitive development, or social interactions. By employing tools like differential equations, phase space analysis, and chaos theory, researchers can capture patterns in how psychological systems change under different conditions. Applications include understanding how individuals adapt to stress, modeling brain activity during cognition, and exploring group behaviors like decision-making or crowd dynamics. These approaches provide a more nuanced view of psychological phenomena, capturing the inherent complexity and variability of human behavior.

Our work in non-linear dynamics extends traditional statistical approaches to capture the complexity of psychological phenomena. Key areas include:

• Development of methods for analyzing psychological time series
• Application of dynamical systems theory to psychological processes
• Study of emergent phenomena in social and cognitive systems
• Integration of chaos theory with psychological measurement
• Advanced modeling of developmental processes

Faculty: Steven Boker

Large Language Models and Artificial Intelligence

Our program is at the forefront of integrating artificial intelligence and large language models (LLMs) into psychological research methods. This cutting-edge area focuses on:

• Development of AI-powered psychological assessment tools
• Validation of LLM-generated psychological instruments in silica
• Investigation of AI models' cognitive architectures
• Application of transformer models to psychological data analysis
• Integration of machine learning with traditional psychometric approaches
• Ethical considerations in AI-assisted psychological research
• Development of interpretable AI models for psychological applications

Our faculty leads innovative projects combining AI with quantitative psychology, including the development of AI-GENIE and other tools that leverage LLMs for psychological measurement. We also investigate how insights from cognitive psychology and psychometrics can inform the development of more robust and interpretable AI systems.

Faculty: Hudson Golino, Steven Boker

The OpenMX Project

The OpenMx Project. Funded by NIH, OpenMx is a software development project for an open source Structural Equation Modeling (SEM) package that is free of charge and tied into the R statistical system. The OpenMx project involves collaborators at the University of Virginia, Medical College of Virginia, University of Chicago, University of Houston, McMaster University and University of Edinburgh. The project was officially released in the fall of 2010 and has been downloaded more than 70,000 times. Statistical researchers, quantitative psychologists, and software developers who may be thinking about writing extensions to SEM are encouraged to visit the OpenMx Wiki where they can learn about the project and possibly either contribute to the project or find software and open source code that may help them in their own work.

EGAnet: Exploratory Graph Analysis: A Framework for Estimating the Number of Dimensions in Multivariate Data Using Network Psychometrics

An implementation of the Exploratory Graph Analysis (EGA) framework for dimensionality assessment. EGA is part of a new area called network psychometrics that focuses on the estimation of undirected network models in psychological datasets. EGA estimates the number of dimensions or factors using graphical lasso or Triangulated Maximally Filtered Graph (TMFG) and a weighted network community analysis. A bootstrap method for verifying the stability of the estimation is also available. The fit of the structure suggested by EGA can be verified using confirmatory factor analysis and a direct way to convert the EGA structure to a confirmatory factor model is also implemented. Documentation and examples are available.

https://r-ega.net

https://cran.r-project.org/web/packages/EGAnet/index.html

transforEmotion: Sentiment Analysis for Text, Image and Video Using Transformer Models

With transforEmotion you can use cutting-edge AI/transformer models for zero-shot emotion classification of text, image, and video in R, all without the need for a GPU, subscriptions, paid services, or using Python. All data is processed locally on your machine, and nothing is sent to any external server or third-party service. This ensures full privacy for your data.

https://github.com/atomashevic/transforEmotion

https://cran.r-project.org/web/packages/transforEmotion/index.html

latentFactoR: Data Simulation Based on Latent Factors

Generates data based on latent factor models. Data can be continuous, polytomous, dichotomous, or mixed. Skews, cross-loadings, wording effects, population errors, and local dependencies can be added. All parameters can be manipulated

https://cran.r-project.org/web/packages/latentFactoR/index.html

AI-GENIE: Generative Psychometrics with Automatic Item Generation and Validation via Network-Integrated Evaluation

The rapid advancement of artificial intelligence (AI), particularly large language models (LLMs), has introduced powerful tools for various research domains, including psychological scale development. This study presents a fully automated method to efficiently generate and select high-quality, non-redundant items for psychological assessments using LLMs and network psychometrics. Our approach called, Automatic Item Generation and Validation via Network-Integrated Evaluation (AI-GENIE), reduces reliance on expert intervention by integrating generative AI with the latest network psychometric techniques. The efficacy of AI-GENIE was evaluated through Monte Carlo simulations using the Mixtral, Gemma 2, Llama 3, GPT 3.5, and GPT 4o models to generate item pools that mimic Big Five personality assessment. The results demonstrated improvement in item selection efficiency, with overall average increases of 9.78-17.80 in normalized mutual information in the final item pool across all models. After, each model in AI-GENIE generated a Big Five inventory that was administered to independent, representative samples (N = 1000 each) in the U.S. The empirical results show that the items produced across all models were diverse, theoretically consistent, and structurally stable. Taken together, these findings demonstrate that AI-GENIE is a highly effective tool to automate and streamline scale development and validation process.

Package to be released soon.

Pre-print here: https://osf.io/preprints/psyarxiv/fgbj4