Esametry: Statistical Methods for Personalized Digital Health
An n-of-1 study is an idiographic, within-individual study of one person's recurrent characteristics and patterns under various exposure or treatment conditions. These exposure periods make up repeating intervals that can be accurately described as a set of partitioned time series of variables (e.g., outcomes and predictors). The core idea is that there are stable periodic or cyclic patterns that are predictable.
Our vision is to unite the various statistical methods focused on studying intra-individual cyclic patterns using digital health data under one statistical field of study. We call this field esametry (pronounced “ee-sa-metry”). This term is derived from isa (pronounced “ee-SA”), the Filipino-Tagalog word for “one”. For more on how we derived this term, see this post.
Grounded in biostatistics and causal inference for n-of-1 studies, but spans many domains of statistics. To date, these may (or actually do) include missing data, longitudinal analysis, functional data analysis, accelerometry, machine learning, series-of-n-of-1 analysis, and meta-analysis. However, the unifying strand that defines this field is its focus on intra-individual cyclic patterns.
Related to periodic pattern analysis or similarity search. However, these are engineering techniques aimed at assessing similarity between multiple time series or their partitions (i.e., they do not focus on statistical estimation and inference). This aspect of esametry is perhaps more closely aligned with the statistics field of functional data analysis.
Most useful when the patterns of association or causal mechanisms of interest are known or suspected to be highly specific to each individual. That is, esametric methods should be used when average associations or mechanisms across individuals are ill-defined or do not exist.
Comprised of technical methods for data collection and analysis, but is not limited in application to just health. We will focus on human individuals for now, but an “individual” need not be a patient—or even a human being. Rather, an “individual” could refer to an athlete, audience member, shopper, body part, animal, vehicle, sports team, organization, institution, country, geographic region, financial instrument, recurrent group behavior, geologic phenomenon, etc.
- To unite methods for personalized digital health under one cohesive field of statistics.
- To optimize cross-disciplinary collaboration in order to improve and maintain each individual's own health.
To advance our Mission, we created this blog to help esametrically inclined statistics professionals connect by posting about their work. To do this, we'll discuss:
- Theory: Statistics and biostatistics, psychometrics, data science, machine learning, bioinformatics, and bioengineering methodologists are encouraged to share their ideas for building or adapting techniques that meet the needs of digital health investigators (e.g., health behavior psychologists and clinicians, chronic disease researchers, n-of-1 trialists, self-trackers, self-researchers).
- Applications: Digital health investigators are encouraged to share their experiences in designing studies and analyzing data. This will help us methodologists hone in on the techniques needed in applications of greatest interest.
Over time, we hope these conversations will help define the field of esametry, and thereby foster sustained development of this field of expanded n-of-1 studies.
- Short Post (#iksi): every Monday or first non-holiday weekday + sporadic
- Long Post (#haba): occasional