pith. sign in

arxiv: 2606.08261 · v1 · pith:ZRAUEM64new · submitted 2026-06-06 · 📊 stat.ME · stat.AP

Sparse Longitudinal Functional Principal Component Analysis for Episodic Ambulatory Behavioral Assessments

Nidhi Pai , Yu Fang , Srijan Sen , Zhenke Wu , Erjia Cui This is my paper

Pith reviewed 2026-06-27 19:15 UTC · model grok-4.3

classification 📊 stat.ME stat.AP
keywords sparse functional data analysislongitudinal functional principal component analysispenalized splinesmental fatigueambulatory assessmenttyping speedIntern Health Study
0
0 comments X

The pith

Sparse LFPCA decomposes variability in episodic typing speed trajectories to reveal new participant- and day-level mental fatigue patterns.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes sparse LFPCA to analyze sparsely observed functional data such as smartphone typing speeds collected episodically as a proxy for mental fatigue. It formulates covariance estimation as a structured penalized spline regression problem so that multiple covariance components can be estimated and smoothed together by borrowing strength across the time domain. Simulations establish that the method recovers eigenfunctions accurately, produces reasonable curve predictions, and performs at least as well as existing approaches. When applied to data from the Intern Health Study, the decomposition identifies distinct participant-specific and day-specific patterns that earlier analyses had not captured. A reader would care because these patterns could guide the timing of just-in-time interventions that improve workplace safety and productivity.

Core claim

Treating typing speed trajectories as sparsely observed functional data and casting covariance estimation as structured penalized spline regression allows sparse LFPCA to decompose variability into eigenfunctions and generate predictions for individual curves, thereby uncovering new and interpretable participant- and day-level patterns in the Intern Health Study typing speed data.

What carries the argument

structured penalized spline regression for simultaneous estimation and smoothing of multiple covariance components while borrowing information across locations in the functional domain

If this is right

  • Simulations confirm accurate eigenfunction estimation and reasonable predictions for underlying curves.
  • The method achieves similar or superior performance compared with existing alternatives.
  • Analysis of the typing speed data identifies new participant- and day-level patterns not captured by prior work.
  • The extracted patterns can be used to tailor behavioral interventions for mental fatigue.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same covariance-borrowing approach could be tested on other episodically sampled smartphone behavioral streams such as activity or response time.
  • If the participant- and day-level components prove stable, they could be fed forward into real-time prediction models that trigger interventions at the individual rather than population level.
  • Extensions that relax the Gaussian assumption or accommodate irregular missingness blocks would be natural next checks on the method's scope.

Load-bearing premise

Typing speed trajectories can be validly treated as sparsely observed functional data whose covariance structure is amenable to simultaneous estimation and smoothing via structured penalized spline regression that borrows information across the functional domain.

What would settle it

A controlled simulation in which the true covariance does not permit borrowing across the domain and the method recovers markedly inaccurate eigenfunctions would falsify the claim.

Figures

Figures reproduced from arXiv: 2606.08261 by Erjia Cui, Nidhi Pai, Srijan Sen, Yu Fang, Zhenke Wu.

Figure 1
Figure 1. Figure 1: Selected participants and days from the IHS typing speed data. Columns cor [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Boxplots of integrated squared error (ISE) for covariance functions in Simulation [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Boxplots of ISEs for eigenfunctions in Simulation 1 (sparse LFPCA). Each quadrant [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Estimated marginal mean functions in the SensorKit data. The left panel shows [PITH_FULL_IMAGE:figures/full_fig_p024_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Estimated eigenfunctions from SensorKit typing speed data. The x-axis is time [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
read the original abstract

Accurately monitoring mental fatigue is critical for improving workplace safety and productivity. A recent study examined unobtrusively collected smartphone typing speed as a potential ambulatory proxy assessment of mental fatigue using data from the Intern Health Study (IHS). While population-level average typing speed patterns were found to be consistent with validated measures of mental fatigue, how these trajectories vary across participants and days may inform opportune moments for just-in-time interventions and remains an open question. Treating typing speed trajectories as sparsely observed functional data, we propose a novel sparse longitudinal functional principal component analysis (sparse LFPCA) method for decomposing variability and predicting individual curves. Specifically, sparse data are accommodated by casting covariance estimation as a structured penalized spline regression problem, enabling simultaneous estimation and smoothing of multiple covariance components while borrowing information across locations in the functional domain. Simulations show that sparse LFPCA (1) accurately estimates eigenfunctions and generates reasonable predictions for underlying curves, and (2) achieves similar or superior performance compared to existing alternatives. Our analysis of typing speed data collected from IHS reveals new and interpretable participant- and day-level patterns not captured by previous analyses and can be used to tailor behavioral interventions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript proposes sparse longitudinal functional principal component analysis (sparse LFPCA) for sparsely observed functional data. Covariance estimation is formulated as a structured penalized spline regression problem that enables simultaneous estimation and smoothing while borrowing information across the functional domain. Simulations demonstrate accurate eigenfunction recovery, reasonable curve predictions, and competitive or superior performance relative to existing methods. The approach is applied to typing speed trajectories from the Intern Health Study, identifying new participant- and day-level patterns not captured in prior analyses.

Significance. If the simulation results and real-data patterns hold, the work supplies a practical extension of FPCA tailored to episodic ambulatory behavioral data. The structured penalized spline formulation addresses sparsity by sharing strength across locations, which is relevant for mental fatigue monitoring via smartphone metrics. The application illustrates utility for extracting interpretable components that could guide just-in-time interventions. Positive elements include simulation coverage across sparsity levels and sensitivity checks on smoothing parameters.

minor comments (3)
  1. [Simulations section] The abstract states that simulations support accurate eigenfunction estimation, but the main text should explicitly report the range of sparsity levels tested and how they align with the observed typing speed data density.
  2. [Application section] The claim that the extracted patterns are 'not captured by previous analyses' would be strengthened by a direct quantitative comparison (e.g., variance explained or prediction error) against the population-level averages reported in the referenced IHS study.
  3. [Methods] Notation for the structured penalty and the borrowing of information across the functional domain should be introduced with a brief equation or diagram in the methods to improve readability for readers unfamiliar with penalized spline covariance models.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our manuscript on sparse LFPCA and for recommending minor revision. The referee's description accurately reflects the method's formulation as structured penalized spline regression, the simulation results, and the application to typing speed trajectories from the Intern Health Study.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper proposes sparse LFPCA via structured penalized spline regression for covariance estimation on sparsely observed functional data. Simulations validate eigenfunction recovery and prediction accuracy against alternatives, while the typing speed application extracts participant- and day-level patterns. No load-bearing step reduces a claimed prediction or eigenfunction to a fitted parameter by construction, nor does any uniqueness theorem or ansatz reduce to self-citation. The derivation chain is self-contained with external simulation benchmarks and sensitivity checks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; full details on free parameters, axioms, and any invented entities cannot be extracted. The method relies on standard assumptions of functional data analysis and penalized splines, but these are not enumerated here.

pith-pipeline@v0.9.1-grok · 5743 in / 1051 out tokens · 18196 ms · 2026-06-27T19:15:34.061846+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

51 extracted references · 3 canonical work pages · 2 internal anchors

  1. [1]

    Journal of the American Statistical Association , volume=

    Functional data analysis for sparse longitudinal data , author=. Journal of the American Statistical Association , volume=. 2005 , publisher=

    2005

  2. [2]

    2005 , publisher=

    Functional data analysis , author=. 2005 , publisher=

    2005

  3. [3]

    Statistics and Computing , volume=

    Fast covariance estimation for sparse functional data , author=. Statistics and Computing , volume=. 2018 , publisher=

    2018

  4. [4]

    Statistics and Computing , volume=

    Fast covariance estimation for high-dimensional functional data , author=. Statistics and Computing , volume=. 2016 , publisher=

    2016

  5. [5]

    Journal of Computational and Graphical Statistics , volume=

    Fast multilevel functional principal component analysis , author=. Journal of Computational and Graphical Statistics , volume=. 2023 , publisher=

    2023

  6. [6]

    The Annals of Applied Statistics , volume=

    Multilevel functional principal component analysis , author=. The Annals of Applied Statistics , volume=

  7. [7]

    Stat , volume=

    Multilevel sparse functional principal component analysis , author=. Stat , volume=. 2014 , publisher=

    2014

  8. [8]

    Electronic Journal of Statistics , volume=

    Longitudinal functional principal component analysis , author=. Electronic Journal of Statistics , volume=

  9. [9]

    Statistical Modelling , volume=

    Functional linear mixed models for irregularly or sparsely sampled data , author=. Statistical Modelling , volume=. 2016 , publisher=

    2016

  10. [10]

    2017 , publisher=

    Introduction to functional data analysis , author=. 2017 , publisher=

    2017

  11. [11]

    2024 , publisher=

    Functional data analysis with R , author=. 2024 , publisher=

    2024

  12. [12]

    Tutorial on

    Jiang, Ziren and Crainiceanu, Ciprian and Cui, Erjia , journal=. Tutorial on. 2025 , publisher=

    2025

  13. [13]

    Journal of Computational and Graphical Statistics , volume=

    Functional additive mixed models , author=. Journal of Computational and Graphical Statistics , volume=. 2015 , publisher=

    2015

  14. [14]

    Statistics in Medicine , volume=

    Fixed-effects inference and tests of correlation for longitudinal functional data , author=. Statistics in Medicine , volume=. 2022 , publisher=

    2022

  15. [15]

    Journal of Computational and Graphical Statistics , volume=

    Fast univariate inference for longitudinal functional models , author=. Journal of Computational and Graphical Statistics , volume=. 2022 , publisher=

    2022

  16. [16]

    The Annals of Applied Statistics , volume=

    Longitudinal high-dimensional principal components analysis with application to diffusion tensor imaging of multiple sclerosis , author=. The Annals of Applied Statistics , volume=

  17. [17]

    Stat , volume=

    Longitudinal functional data analysis , author=. Stat , volume=. 2015 , publisher=

    2015

  18. [18]

    Biostatistics , volume=

    Bayesian analysis of longitudinal and multidimensional functional data , author=. Biostatistics , volume=. 2022 , publisher=

    2022

  19. [19]

    Biostatistics , volume=

    Hybrid principal components analysis for region-referenced longitudinal functional EEG data , author=. Biostatistics , volume=. 2020 , publisher=

    2020

  20. [20]

    Journal of the American Statistical Association , volume=

    Regression analysis of asynchronous longitudinal functional and scalar data , author=. Journal of the American Statistical Association , volume=. 2022 , publisher=

    2022

  21. [21]

    Biometrics , volume=

    Modeling longitudinal skewed functional data , author=. Biometrics , volume=. 2024 , publisher=

    2024

  22. [22]

    Fast Penalized Generalized Estimating Equations for Large Longitudinal Functional Datasets

    Fast Penalized Generalized Estimating Equations for Large Longitudinal Functional Datasets , author=. arXiv:2506.20437 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  23. [23]

    Biometrics , volume=

    Structured functional principal component analysis , author=. Biometrics , volume=. 2015 , publisher=

    2015

  24. [24]

    Journal of Applied Statistics , volume=

    Functional principal component models for sparse and irregularly spaced data by Bayesian inference , author=. Journal of Applied Statistics , volume=. 2024 , publisher=

    2024

  25. [25]

    Stat , volume=

    Fast covariance estimation for multivariate sparse functional data , author=. Stat , volume=. 2020 , publisher=

    2020

  26. [26]

    Bayesian Multivariate Sparse Functional Principal Components Analysis

    Bayesian Multivariate Sparse Functional PCA , author=. arXiv preprint arXiv:2509.03512 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  27. [27]

    Journal of Multivariate Analysis , volume=

    Robust functional principal component analysis for non-Gaussian longitudinal data , author=. Journal of Multivariate Analysis , volume=. 2022 , publisher=

    2022

  28. [28]

    arXiv preprint arXiv:2503.21913 , year=

    A novel smoothing-based goodness-of-fit test of covariance for multivariate sparse functional data , author=. arXiv preprint arXiv:2503.21913 , year=

    work page arXiv

  29. [29]

    Biometrika , volume=

    Principal component models for sparse functional data , author=. Biometrika , volume=. 2000 , publisher=

    2000

  30. [30]

    Flexible smoothing with

    Eilers, Paul HC and Marx, Brian D , journal=. Flexible smoothing with. 1996 , publisher=

    1996

  31. [31]

    Chemometrics and Intelligent Laboratory Systems , volume=

    Multivariate calibration with temperature interaction using two-dimensional penalized signal regression , author=. Chemometrics and Intelligent Laboratory Systems , volume=. 2003 , publisher=

    2003

  32. [32]

    2012 , journal=

    Asymptotic optimality and efficient computation of the leave-subject-out cross-validation , author=. 2012 , journal=

    2012

  33. [33]

    2017 , publisher=

    Generalized additive models: an introduction with R , author=. 2017 , publisher=

    2017

  34. [34]

    Journal of the American Statistical Association , volume=

    A note on penalized spline smoothing with correlated errors , author=. Journal of the American Statistical Association , volume=. 2007 , publisher=

    2007

  35. [35]

    Journal of the Royal Statistical Society: Series B (Methodological) , volume=

    Estimating the mean and covariance structure nonparametrically when the data are curves , author=. Journal of the Royal Statistical Society: Series B (Methodological) , volume=. 1991 , publisher=

    1991

  36. [36]

    Biostatistics , volume=

    Fast methods for spatially correlated multilevel functional data , author=. Biostatistics , volume=. 2010 , publisher=

    2010

  37. [37]

    Journal of the American Statistical Association , volume=

    Generalized multilevel functional regression , author=. Journal of the American Statistical Association , volume=. 2009 , publisher=

    2009

  38. [38]

    2008 , publisher=

    A matrix handbook for statisticians , author=. 2008 , publisher=

    2008

  39. [39]

    Karhunen, Kari , journal=

  40. [40]

    Kosambi, D. D. , title =. Journal of the Indian Mathematical Society , volume =

  41. [41]

    Fonctions al\'

    Loeve, Michel , journal=. Fonctions al\'. 1948 , publisher=

    1948

  42. [42]

    Philosophical Transactions of the Royal Society , volume=

    Functions of positive and negative type, and their connection with the theory of integral equations , author=. Philosophical Transactions of the Royal Society , volume=. 1909 , publisher=

    1909

  43. [43]

    Biometrics , pages=

    Best linear unbiased estimation and prediction under a selection model , author=. Biometrics , pages=. 1975 , publisher=

    1975

  44. [44]

    Biometrika , volume=

    On a formula for the product-moment coefficient of any order of a normal frequency distribution in any number of variables , author=. Biometrika , volume=. 1918 , publisher=

    1918

  45. [45]

    PLOS Digital Health , volume=

    Patterns of smartphone typing performance by time awake: implications for unobtrusive ambulatory mental fatigue assessment , author=. PLOS Digital Health , volume=. 2026 , publisher=

    2026

  46. [46]

    Feasibility and acceptability of collecting passive phone usage and sensor data via

    Funk, Courtney and Zhao, Zhuo and Horwitz, Adam G and Fang, Yu and Pereira-Lima, Karina and Kheterpal, Vik and Sen, Srijan and Frank, Elena , journal=. Feasibility and acceptability of collecting passive phone usage and sensor data via. 2025 , publisher=

    2025

  47. [47]

    SensorKit , howpublished =

  48. [48]

    Exploring the potential of

    Langholm, Carsten and Kowatsch, Tobias and Bucci, Sandra and Cipriani, Andrea and Torous, John , journal=. Exploring the potential of. 2023 , publisher=

    2023

  49. [49]

    Fatigue in the

    Ricci, Judith A and Chee, Elsbeth and Lorandeau, Amy L and Berger, Jan , journal=. Fatigue in the. 2007 , publisher=

    2007

  50. [50]

    Archives of Surgery , volume=

    Surgeon fatigue: a prospective analysis of the incidence, risk, and intervals of predicted fatigue-related impairment in residents , author=. Archives of Surgery , volume=. 2012 , publisher=

    2012

  51. [51]

    Statistics in medicine , volume=

    Using simulation studies to evaluate statistical methods , author=. Statistics in medicine , volume=. 2019 , publisher=

    2019