Characterizing Daily Affective Dynamics in a Single Individual

A control-theoretic N-of-1 study with ecological momentary assessment.

A 70-day single-subject self-experiment that tests whether nightly melatonin predicts daily affective state, modeling the individual as a dynamical system and treating melatonin as a probe of that system rather than as a candidate treatment. Internal behavioral drivers (perceived agency and metacognitive awareness) jointly explained 51.5% of daily-mood variance beyond an AR(1) baseline, whereas the melatonin probe added 0.01% and produced only a non-significant trend toward lower day-to-day variability (α₁ = −1.11, p = .071), pointing to a stabilizing rather than a mood-lifting effect.

Author: Miura Meng · meng10@upenn.edu · ORCID 0009-0004-1522-1997

Keywords: N-of-1; ecological momentary assessment; dynamical systems; autoregressive model; idiographic; personal science; intensive longitudinal data; affect dynamics

Manuscript

All versions are in paper/ (latest: v9).

Repository layout

• randomization/: pre-registered 70-day schedule (schedule.json) and protocol
• data/: cleaned EMA, observation-level (miura_ema_70day.csv, n = 195) and daily (miura_ema_70day_daily.csv, n = 70)
• src/data_logger.py: iOS Shortcuts entry validation
• src/analysis/: analysis scripts 01–09 (run order below)
• outputs/: cleaned frames, model tables, and results_table.csv (every number reported in the paper)
• figures/: figures 1–7
• paper/: manuscript versions

Reproducing the analysis

Python 3.10+ (pip install -r requirements.txt), run from the repository root:

python src/analysis/01_data_prep.py                    # clean and verify the dataset
python src/analysis/02_ar1_models.py                   # nested AR(1) models + incremental R²
python src/analysis/02b_ar1_obslevel.py                # observation-level AR(1) + half-life
python src/analysis/03_state_space.py                  # local-level Kalman filter
python src/analysis/04_metacontrol.py                  # melatonin × metacognition interaction
python src/analysis/05_variability_and_interactions.py # variability, interactions, state-dependence
python src/analysis/08_results_table.py                # assemble results_table.csv
python src/analysis/06_make_figures.py                 # regenerate figures 1–7

Two supplementary robustness analyses run in R (≥ 4.1):

source("src/analysis/07_bayesian_robustness.R")    # Bayesian M2 (brms)        -> Table 2 / §3.6
source("src/analysis/09_multilevel_robustness.R")  # mixed-effects (lme4)      -> Table 3 / §3.7

Every number in the manuscript can be recovered from outputs/. The fig1–fig4 PNGs are the author's polished originals; re-running 06_make_figures.py reproduces the same content in matplotlib with minor stylistic differences.

Method in brief

A 70-day alternating-treatment design randomized each day to melatonin or control (no run longer than two consecutive days). EMA was collected three times daily (~10:00, 16:00, 22:00) via iOS Shortcuts: mood, agency, and metacognition on 0–100 sliders plus a melatonin indicator. Agency and metacognition were added on Day 18, so analyses using them cover Days 18–70. The final sample is 195 observations across 70 study days (92.9% compliance), with 35 control and 35 melatonin days.

Availability and citation

Released under CC-BY 4.0.

Meng, M. (2026). Characterizing daily affective dynamics in a single individual: A control-theoretic N-of-1 study with ecological momentary assessment. Manuscript in preparation. https://github.com/haomeng797-ship-it/N-of-1-Melatonin-Study