Eigenmode Method for Psychotherapeutic Research

Emma T. Swan
April, 2026

Valid Heterogeneous Clinical Trials

Introduction

This pipeline aims to assess psychotherapeutic models using analytic methods that accommodate flexible, individualized delivery without sacrificing statistical rigor.

Video

Research Question

Do the modality-delivery variables (time, interventions, context variables, and psychological measures) form a stable eigenmode structure with a dominant eigenvalue that is invariant across technique groups, such that any specific technique demonstrates the same underlying pattern in vector space of the psychotherapeutic modality?

CC 4.0

Example Dataset

We'll use an existing dataset as an example of how this proposed analysis pipeline works.

The data set is from a study on positive psychology that offered one of four techniques found within positive psychology to the 202 subjects. Happiness (AHI) and depressions (CESD) scores before and after the intervention were collected. For PCA analysis I extracted these two testing occasions, splitting the column of each score into pre and post scores so that the timing would be included in the PCA.

J. Woodworth, Rosalind; O'Brien-Malone, Angela; Diamond, Mark R.; Schüz, Benjamin (2018). A randomized placebo-controlled trial of positive psychology interventions in Australia. figshare. Dataset. https://doi.org/10.6084/m9.figshare.1577563.v1

Video

Analysis Methodology

Pipeline Tutorial

Log & Robust Scaling
Preliminary AnalysisPrincipal Component Analysis (PCA)
ClusteringModeling

Example Results

Video

The results overall indicate a low-dimensional eigenmode structure with a significant dominant principal component comprised of a wellbeing axis of positive happiness and negative depression distributed and predictive according to subject baseline rather than change across the entire subject pool. The similarities in the principal component structure of each technique group corresponds to a consistent modality structure regardless of technique used.

Future Considerations

The analysis pipeline largely answered the research question with the indication of the underlying and predictive pattern consistent across techniques.

Further refinements could include incorporating physiological and context data in addition to technique and psychological results data, using non-clinical surveys to reduce the skew found in the CESD, testing the same subjects across all techniques, and cross-validation.

Next steps could be to develop a data collection template and a research design based on this analysis pipeline to limit dataset modifications and to ensure all the relevant data is collected to more completely answer the research question.