Individualized Analytics: Within-Person and Idiographic Analysis

Mohammed Saqr, Hibiki Ito, Sonsoles López-PernasOpenalex Unknown 2025 · DOI · OpenAlex

Abstract

Two research paradigms are commonly used: between-person (often known as nomothetic or group-based) and within-person analysis. The between-person approach captures data across different people and gathers their variations. Within-person research captures data within the same person, gathering these variations across times or occasions. These two approaches converge if and only if ergodicity holds. Ergodicity claims that people are similar (so that it does not matter whom to measure) and they do not change over time (so that it does not matter when to measure). However, this is often not the case in education research. Thus, the within-person approach is needed. When within-person analysis is conducted on a single person (N=1), it is known as idiographic analysis. In this chapter, we describe the main characteristics of each approach, as well as common data sources and methods used in within-person and idiographic analysis.

Introduction: The Generalizability Crisis in Learning Analytics

Most current learning analytics systems rely on "one-size-fits-all" group averages. While these nomothetic approaches are useful for identifying broad trends, they often fail to predict or support individual students effectively. This paper advocates for a paradigm shift toward idiographic analysis—modeling the unique, time-varying data of a single person ($N=1$) to provide truly personalized educational interventions.

Traditionally, researchers assume ergodicity: the idea that everyone in a group is similar (homogeneity) and that their behavior remains stable over time (stationarity). When these assumptions fail, applying group-level insights to individuals leads to the Ecological Fallacy—assuming what is true for the group is true for every member.

Comparison of Between-person, Within-person, and Idiographic research paradigms.

As shown in the figure above, the between-person approach (A) looks across individuals at a single point in time ($t_0$), while the within-person approach (B) tracks multiple individuals over time. Idiographic analysis (C) takes this further by generating unique models for each individual, recognizing that the relationships between variables may differ fundamentally from person to person.

Methods for Intensive Longitudinal Data

To handle the complexity of "intensive longitudinal data" (many observations of the same person over time), several advanced statistical and computational methods are employed:

  • P-technique: Applies factor analysis to a single individual’s time-series data to identify unique latent constructs, such as a specific student's motivation profile.
  • Graphical Vector Autoregression (VAR): A network-based method capturing temporal effects (how past states influence the present) and contemporaneous effects (how different variables co-occur).
  • Sequence Mining: Identifies patterns in digital logs (e.g., LMS clicks) to uncover unique learning "tactics" or "trajectories."
  • Idiographic Machine Learning: Training models on an individual's own data allows "feature importance" to vary; for one student, sleep might predict success, while for another, social interaction is key.
  • GIMME (Group Iterative Multiple Model Estimation): A hybrid approach that identifies paths common to a group while allowing for person-specific variations.

Key Insights: The Myth of the "Average Person"

A central finding is that the "average person" is often a statistical fiction. In many studies, the feature importance of a group-level model did not match the feature importance of any single individual in that group.

Individual student trajectories vs. the class mean, illustrating the myth of the average student.

This phenomenon is clearly illustrated in the figure above. While the "Class Mean" shows a steady upward trend, individual students (A through E) follow wildly different paths. Some improve significantly faster than the mean, while others actually decline. A system based solely on the class mean would fail to intervene for Student B or E, whose performance is dropping despite the positive group trend. This is a classic example of Simpson’s Paradox, where group-level correlations can be the exact opposite of individual trends.

Results and Practical Advantages

Quantitative findings from the literature support the superiority of individualized models:

  • Better Prediction: One study of 286 students found that a "within-between hybrid model" significantly outperformed standard group-level models in predicting final marks.
  • Heterogeneity in Stress: Smartphone log research showed that while sleep and social media use predicted stress at the group level, these were not predictive for a large subgroup of individuals driven by different factors.
  • Fairness and Privacy: Idiographic approaches bypass majority bias by treating every student as their own baseline—crucial for students with unique needs. Furthermore, person-specific models could be processed locally on devices, enhancing data privacy.

Limitations and Future Outlook

Despite its promise, idiographic analysis faces challenges. It requires a high volume of "micro-level" data, which can lead to participant fatigue. The methods are also statistically complex. However, the rise of wearable devices and automated machine learning (AutoML) offers a path to scaling these $N=1$ solutions, moving education toward a truly personalized future.

Summary updated

Abstract

Abstract Two research paradigms are commonly used: between-person (often known as nomothetic or group-based) and within-person analysis. The between-person approach captures data across different people and gathers their variations, hence the name of between-person. Within-person research captures data within the same person, gathering these variations across times or occasions and thus the name. These two approaches converge if and only if ergodicity holds. Ergodicity claims that people are similar (so that it does not matter whom to measure) and they do not change over time (so that it does not matter when to measure). However, this is often not the case in education research. Thus, the within-person approach is needed. When within-person analysis is conducted on a single person (N=1), it is known as idiographic analysis. In this chapter, we describe the main characteristics of each approach, as well as common data sources and methods used in within-person and idiographic analysis.
BibTeX
@inproceedings{saqr2025individualized,
  author = {Mohammed Saqr and Hibiki Ito and Sonsoles López-Pernas},
  title = {{Individualized Analytics: Within-Person and Idiographic Analysis}},
  booktitle = {Openalex Unknown},
  year = {2025},
  doi = {10.1007/978-3-031-95365-1\_18},
}
Figure