Research

My research addresses fundamental questions about educational measurement and assessment in small-sample, many-group contexts:

How can we design assessments and models that reliably measure student outcomes and support equitable decision-making when data are limited and highly segmented?

This question sits at the heart of my work and connects issues of psychometric modeling, predictive analytics, and educational policy.

Research Areas

Integrating Machine Learning and Psychometrics

I bring neural networks, graph-based methods, and other machine learning tools into psychometric modeling, from detecting item bias to estimating parameters and scoring.

Intersectional DIF
Neural network approaches, including multi-label classification, for detecting differential item functioning across intersecting demographic groups with small samples.
Machine Learning for Prediction
Machine learning for binary prediction with incomplete, imbalanced data and non-linear effects.
LLM-Generated Items
Evaluating AI-generated assessment and survey items against human-written items.
Psychometric and Computational Integration
Combining psychometric and computational approaches to support equity, interpretability, and generalizability.

In progress: Extending neural network approaches to item bias detection so they generalize to new settings.


Cognitive Diagnostic Models

I develop methodological frameworks that make diagnostic models practical, from design and specification to model fit, so they give reliable, actionable feedback.

Practical Recommendations
Sample size, test length, item fit, and Q-matrix recommendations for cognitive diagnostic models in practical applications.
Methodological Review
Reviewing recent methodological advances that address practical challenges in cognitive diagnostic models.

In progress: Developing a probabilistic approach to Q-matrix estimation.


Measurement in Complex Data

I develop methods for settings that challenge standard models: small samples, many groups, sparse responses, and missing data.

Many-Group Measurement Invariance
Graph-theoretic methods, such as transitive DIF clustering, for identifying partial measurement invariance and improving parameter estimation across many groups.
Item Response Theory
Multiple-group and higher-order IRT models, including calibration with structural missingness.

In progress: How decisions about sparse response categories affect item bias detection.


Applied Impact

In my applied work, I collaborate with K–12 schools and community partners to study barriers to student success, and I apply psychometric and statistical methods in healthcare and psychological research.

Studying student belonging and chronic absenteeism with K–12 and community partners.
Designing dashboards and visualizations to guide equitable resource allocation.
Consulting on research design, statistical analysis, and evidence-based decision-making.

Collaborations & Research Experience