pith. sign in

arxiv: 2509.03512 · v3 · submitted 2025-09-03 · 📊 stat.ME

Bayesian Multivariate Sparse Functional Principal Components Analysis

Joseph Sartini , Scott Zeger , Ciprian Crainiceanu This is my paper

Pith reviewed 2026-05-18 19:05 UTC · model grok-4.3

classification 📊 stat.ME
keywords Bayesian FPCAmultivariate sparse functional dataprincipal component analysisfunctional data analysisspline representationProcrustes alignmentposterior uncertaintychild growth trajectories
0
0 comments X

The pith

MSFAST is a Bayesian method that estimates principal components for multivariate sparse functional data while accounting for uncertainty in the components.

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

The authors develop a fully Bayesian approach called MSFAST for analyzing multivariate functional data that are observed only at sparse points. Traditional frequentist methods estimate the principal components first and then proceed as if those estimates are known without error. MSFAST instead samples the principal components from their posterior distribution, allowing uncertainty to be carried through to all inferences and predictions. This matters for applications like tracking child growth across several body measurements, where data are often sparse and noisy. The method builds on prior work for univariate dense data by adding steps for standardization, basis choice, alignment, and parallel computation to make the multivariate sparse case feasible.

Core claim

MSFAST represents the principal components using orthonormal splines and samples the spline coefficients via parameter expansion in a Bayesian model tailored to multivariate sparse observations. It incorporates covariate standardization to handle scale differences, a suitable orthogonal basis, parameter updates for stability, multi-core acceleration, Procrustes alignment of posterior samples, and efficient prediction. This framework produces valid inferences that reflect uncertainty in the principal components and yields accurate estimates particularly when the signal-to-noise ratio is small.

What carries the argument

The Bayesian hierarchical model in MSFAST that explicitly represents principal components as orthonormal spline expansions and uses parameter expansion to sample from their posterior while aligning samples via Procrustes methods for the multivariate sparse setting.

Load-bearing premise

The success of the method depends on standardization of each functional covariate being sufficient to fix posterior conditioning problems caused by differing scales, together with the orthogonal spline basis and Procrustes alignment delivering stable posterior samples of the principal components.

What would settle it

A simulation experiment with known true principal components and low signal-to-noise ratio in which the MSFAST posterior credible intervals exhibit coverage substantially below the stated level would indicate that the inferences are not uniquely valid.

Figures

Figures reproduced from arXiv: 2509.03512 by Ciprian Crainiceanu, Joseph Sartini, Scott Zeger.

Figure 1
Figure 1. Figure 1: A) Boxplots of RISE and B) kernel smoother of 95% interval coverage probabilities of the underlying smooth functions for MSFAST, mFACEs, mFPCA, and VMP. Columns correspond to covariate and rows to expected number of observations (5 then 10). where ISE(p) Mean is ISE(p) when the subject-specific trajectories are estimated by the corresponding sample means (Yb(p) i (t) = 1 J (p) i P t∈T (p) i Y (p) i (t)). T… view at source ↗
Figure 2
Figure 2. Figure 2: ISE for MSFAST, mFACEs, mFPCA, and VMP. Columns [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Computation time (y-axis in minutes) as a function of number of subjects [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: A) Boxplots of RISE and B) kernel smoother of 95% interval coverage probabilities of the underlying smooth functions for MSFAST, GFSR, VMP, Conditional, FACE, and PACE. Within each panel, columns correspond to SNR (2 then 5), and rows to expected number of observations (5 then 10). We compare MSFAST with: (1) Bayesian generalized sparse FPCA (“GFSR”) by Gertheiss et al. [2017]; (2) variational Bayes using … view at source ↗
Figure 5
Figure 5. Figure 5: A) Length and weight observation times for each CONTENT participant stratified by sex. Y-axis: participant; x-axis: age in days. Highlighted points correspond to the data in Panel B. B) Data from two randomly chosen participants within each gender, color-coded by measure. Y-axis: z-score; x-axis: age in days. in children less than 2 years of age [Jaganath et al., 2014, Crainiceanu et al., 2024a]. The study… view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of MSFAST posterior inference on the FPCs with the estimates provided [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Observed data (points), dynamic predictions from MSFAST (solid lines), and [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Proportion of variance explained in the original data for each truncation number of FPCs [PITH_FULL_IMAGE:figures/full_fig_p026_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The left panel visualizes the elements of a Q = 5 dimensional orthogonalized B-spline basis, as presented by Redd [2012]. The right panel provides the corresponding Q = 5 dimensional Splinet basis as proposed by Liu et al. [2020], suitably augmented with orthogonalized slope and intercept. In both panels, color indicates basis element. From Supplemental [PITH_FULL_IMAGE:figures/full_fig_p035_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Kernel smoother of 95% interval coverage probabilities of the true FPC and mean [PITH_FULL_IMAGE:figures/full_fig_p039_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Computation time (y-axis in minutes) as a function of number of subjects [PITH_FULL_IMAGE:figures/full_fig_p040_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: ISE for MSFAST, GFSR, VMP, Conditional, FACE, and PACE. Columns [PITH_FULL_IMAGE:figures/full_fig_p041_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Kernel smoother of 95% interval coverage probabilities of the true FPC and mean [PITH_FULL_IMAGE:figures/full_fig_p042_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Computation time (y-axis in minutes) as a function of number of participants [PITH_FULL_IMAGE:figures/full_fig_p043_14.png] view at source ↗
read the original abstract

Functional Principal Components Analysis (FPCA) provides a parsimonious, semi-parametric model for multivariate, sparsely-observed functional data. Frequentist FPCA approaches estimate principal components (PCs) from the data, then condition on these estimates in subsequent analyses. As an alternative, we propose a fully-Bayesian inferential framework for multivariate, sparse functional data (MSFAST) which explicitly models the PCs and incorporates their uncertainty. MSFAST builds upon the FAST approach to FPCA for univariate, densely-observed functional data. Like FAST, MSFAST represents PCs using orthonormal splines and samples the orthonormal spline coefficients using parameter expansion. MSFAST extends FAST to multivariate, sparsely-observed data by (1) standardizing each functional covariate to mitigate poor posterior conditioning due to disparate scales; (2) using a better-suited orthogonal spline basis; (3) updating parameterizations for computational stability; (4) introducing routines that leverage multiple cores and threads to accelerate compute; (5) using a Procrustes-based posterior PC alignment procedure; and (6) providing efficient prediction routines. We evaluate MSFAST alongside existing implementations using simulations. MSFAST produces uniquely valid inferences and accurate estimates, particularly in smaller signal-to-noise regimes. MSFAST is motivated by and applied to a study of child growth, with an accompanying vignette illustrating the implementation step-by-step.

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

3 major / 2 minor

Summary. The manuscript proposes MSFAST, a fully Bayesian extension of the FAST method for functional principal components analysis (FPCA) applied to multivariate, sparsely observed functional data. It explicitly models the principal components (PCs) using orthonormal splines and parameter expansion, incorporates their posterior uncertainty, and introduces six extensions: per-covariate standardization, an improved orthogonal spline basis, updated parameterizations for stability, parallel computing routines, a Procrustes-based posterior alignment procedure, and efficient prediction methods. Simulations are claimed to show that MSFAST yields uniquely valid inferences and more accurate estimates than existing approaches, especially in low signal-to-noise regimes; the method is illustrated on a child growth study.

Significance. If the central claims hold, this would represent a useful methodological advance in functional data analysis by supplying a coherent Bayesian framework that propagates uncertainty from PC estimation into downstream inferences, an aspect often ignored in frequentist FPCA. The emphasis on computational scalability via parallelization and prediction routines, together with the explicit handling of multivariate sparse sampling, addresses practical needs in applications such as longitudinal biomedical studies.

major comments (3)
  1. [§3] §3 (Model extensions), paragraph on standardization: the assertion that standardizing each functional covariate 'sufficiently mitigates poor posterior conditioning due to disparate scales' is presented without a supporting sensitivity analysis or theoretical bound; when covariates retain residual scale differences or exhibit highly irregular sparse sampling, the joint posterior for the multivariate orthonormal spline coefficients may remain ill-conditioned, directly undermining the claim of stable and identifiable posterior PCs.
  2. [§4.2] §4.2 (Posterior alignment), Procrustes procedure: the alignment step is introduced to enforce rotational stability across posterior draws, yet no diagnostic is reported (e.g., pre- versus post-alignment trace plots of the leading eigenvalues or cross-covariate correlations) to verify that the procedure does not distort the joint posterior geometry; without such checks the uncertainty quantification central to the 'uniquely valid inferences' claim cannot be confirmed.
  3. [§5] §5 (Simulation study): the abstract states that MSFAST produces 'accurate estimates, particularly in smaller signal-to-noise regimes,' but the reported results lack quantitative metrics (bias, coverage, or RMSE with error bars) and explicit exclusion criteria for low-SNR settings; this absence prevents assessment of whether the claimed gains over frequentist FPCA are load-bearing or merely suggestive.
minor comments (2)
  1. [§2] The notation for the multivariate spline coefficients and the precise form of the parameter-expansion prior could be clarified with an explicit equation reference in the model section.
  2. [§5] Figure captions for the simulation results should include the exact SNR values and sample sizes used in each panel to improve reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped us identify opportunities to strengthen the manuscript. We address each major comment below and describe the revisions we plan to make in response.

read point-by-point responses

  1. Referee: [§3] §3 (Model extensions), paragraph on standardization: the assertion that standardizing each functional covariate 'sufficiently mitigates poor posterior conditioning due to disparate scales' is presented without a supporting sensitivity analysis or theoretical bound; when covariates retain residual scale differences or exhibit highly irregular sparse sampling, the joint posterior for the multivariate orthonormal spline coefficients may remain ill-conditioned, directly undermining the claim of stable and identifiable posterior PCs.

    Authors: We appreciate the referee's observation. The per-covariate standardization is motivated by standard practices for handling scale differences in multivariate settings and is intended to improve posterior conditioning in the orthonormal spline coefficient model. We agree that the current presentation would be strengthened by empirical support. In the revised manuscript we will add a sensitivity analysis that examines the condition number of the posterior covariance matrix (or equivalent diagnostics) across a range of residual scale disparities and irregular sparse sampling patterns, both with and without standardization. revision: yes

  2. Referee: [§4.2] §4.2 (Posterior alignment), Procrustes procedure: the alignment step is introduced to enforce rotational stability across posterior draws, yet no diagnostic is reported (e.g., pre- versus post-alignment trace plots of the leading eigenvalues or cross-covariate correlations) to verify that the procedure does not distort the joint posterior geometry; without such checks the uncertainty quantification central to the 'uniquely valid inferences' claim cannot be confirmed.

    Authors: We thank the referee for highlighting the need for verification of the Procrustes alignment. The procedure is a standard orthogonal transformation used to resolve label switching due to rotational invariance in PCA. To confirm that it does not materially alter the joint posterior geometry, we will include diagnostic comparisons (pre- versus post-alignment trace plots of the leading eigenvalues and cross-covariate correlations) in the revised manuscript or supplementary material. revision: yes

  3. Referee: [§5] §5 (Simulation study): the abstract states that MSFAST produces 'accurate estimates, particularly in smaller signal-to-noise regimes,' but the reported results lack quantitative metrics (bias, coverage, or RMSE with error bars) and explicit exclusion criteria for low-SNR settings; this absence prevents assessment of whether the claimed gains over frequentist FPCA are load-bearing or merely suggestive.

    Authors: We acknowledge that while the simulation figures illustrate performance differences, explicit numerical summaries would facilitate direct assessment. In the revision we will add a table reporting bias, RMSE, and coverage probabilities (with standard errors) for the principal component estimates and downstream quantities, and we will explicitly state the SNR thresholds and any exclusion rules used to define the low signal-to-noise regimes. revision: yes

Circularity Check

0 steps flagged

No significant circularity; MSFAST extends prior FAST with independent modeling and alignment steps evaluated via simulation

full rationale

The paper proposes MSFAST as a new fully Bayesian framework that explicitly adds six extensions to the earlier FAST method, including per-covariate standardization, an improved orthogonal spline basis, Procrustes posterior alignment, and efficient prediction routines. These are presented as new parameterizations and procedures rather than derivations that reduce to the inputs by construction. Central claims of uniquely valid inferences and improved accuracy in low-SNR regimes are assessed through simulation comparisons, not through self-referential definitions or load-bearing self-citations that would force the result. The derivation chain therefore introduces independent content and remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The abstract indicates reliance on orthonormal splines and parameter expansion from prior FAST work, plus modeling assumptions for standardization and posterior alignment; no explicit free parameters or invented entities are named.

axioms (2)
  • domain assumption Orthonormal splines can represent the principal components with parameter expansion allowing sampling of coefficients.
    Invoked when extending FAST to the multivariate sparse setting.
  • ad hoc to paper Standardization of each functional covariate mitigates poor posterior conditioning from disparate scales.
    Listed as one of the six specific extensions for the multivariate case.

pith-pipeline@v0.9.0 · 5766 in / 1448 out tokens · 54309 ms · 2026-05-18T19:05:37.671835+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Sufficient conditions for proper posteriors in fully-Bayesian Functional PCA

    stat.ME 2026-04 unverdicted novelty 5.0

    No additional conditions beyond the spline projection and mixed-effects equivalence are needed for the smoothing prior and posterior to be proper in fully-Bayesian FPCA.

Reference graph

Works this paper leans on

48 extracted references · 48 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    A Conceptual Introduction to Hamiltonian Monte Carlo

    M. Betancourt. A Conceptual Introduction to Hamiltonian Monte Carlo , July 2018. URL http://arxiv.org/abs/1701.02434. arXiv:1701.02434 [stat]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  2. [2]

    Hoffman, Daniel Lee, Ben Goodrich, Michael Betancourt, Marcus Brubaker, Jiqiang Guo, Peter Li, and Allen Riddell

    B. Carpenter, A. Gelman, M. D. Hoffman, D. Lee, B. Goodrich, M. Betancourt, M. Brubaker, J. Guo, P. Li, and A. Riddell. Stan: A Probabilistic Programming Language . Journal of Statistical Software, 76: 0 1--32, Jan. 2017. ISSN 1548-7660. doi:10.18637/jss.v076.i01. URL https://doi.org/10.18637/jss.v076.i01

    work page doi:10.18637/jss.v076.i01 2017

  3. [3]

    Checkley, L

    W. Checkley, L. D. Epstein, R. H. Gilman, R. E. Black, L. Cabrera, and C. R. Sterling. Effects of Cryptosporidium parvum infection in Peruvian children: growth faltering and subsequent catch-up growth. American Journal of Epidemiology, 148 0 (5): 0 497--506, Sept. 1998. ISSN 0002-9262. doi:10.1093/oxfordjournals.aje.a009675

    work page doi:10.1093/oxfordjournals.aje.a009675 1998

  4. [4]

    Checkley, L

    W. Checkley, L. D. Epstein, R. H. Gilman, L. Cabrera, and R. E. Black. Effects of acute diarrhea on linear growth in Peruvian children. American Journal of Epidemiology, 157 0 (2): 0 166--175, Jan. 2003. ISSN 0002-9262. doi:10.1093/aje/kwf179

    work page doi:10.1093/aje/kwf179 2003

  5. [5]

    Y. Chikuse. Statistics on Special Manifolds , volume 174 of Lecture Notes in Statistics . Springer, New York, NY, 2003. ISBN 978-0-387-00160-9 978-0-387-21540-2. doi:10.1007/978-0-387-21540-2. URL http://link.springer.com/10.1007/978-0-387-21540-2. Edited by Bickel, P. and Diggle, P. and Fienberg, S. and Krickeberg, K. and Olkin, I. and Wermuth, N. and Zeger, S

    work page doi:10.1007/978-0-387-21540-2 2003

  6. [6]

    C. M. Crainiceanu and A. J. Goldsmith. Bayesian Functional Data Analysis Using WinBUGS . Journal of Statistical Software, 32: 0 1--33, Jan. 2010. ISSN 1548-7660. doi:10.18637/jss.v032.i11. URL https://doi.org/10.18637/jss.v032.i11

    work page doi:10.18637/jss.v032.i11 2010

  7. [7]

    C. M. Crainiceanu, J. Goldsmith, A. Leroux, and E. Cui. Functional Data Analysis with R . Chapman and Hall/CRC, 2024 a

    work page 2024

  8. [8]

    C. M. Crainiceanu, J. Goldsmith, A. Leroux, and E. Cui. Functional Data Analysis with R , chapter 3. Chapman and Hall/CRC, 2024 b

    work page 2024

  9. [9]

    Craven and G

    P. Craven and G. Wahba. Smoothing noisy data with spline functions. Numerische Mathematik, 1: 0 377--403, 1979

    work page 1979

  10. [10]

    Simulated

    A. Gelman and D. B. Rubin. Inference from Iterative Simulation Using Multiple Sequences . Statistical Science, 7 0 (4): 0 457--472, Nov. 1992. ISSN 0883-4237, 2168-8745. doi:10.1214/ss/1177011136. URL https://projecteuclid.org/journals/statistical-science/volume-7/issue-4/Inference-from-Iterative-Simulation-Using-Multiple-Sequences/10.1214/ss/1177011136.f...

    work page doi:10.1214/ss/1177011136 1992

  11. [11]

    Gertheiss, J

    J. Gertheiss, J. Goldsmith, and A.-M. Staicu. A note on modeling sparse exponential-family functional response curves. Computational Statistics & Data Analysis, 105: 0 46--52, Jan. 2017. ISSN 0167-9473. doi:10.1016/j.csda.2016.07.010. URL https://www.sciencedirect.com/science/article/pii/S0167947316301748

    work page doi:10.1016/j.csda.2016.07.010 2017

  12. [12]

    Goldsmith, F

    J. Goldsmith, F. Scheipl, L. Huang, J. Wrobel, C. Di, J. Gellar, J. Harezlak, M. W. McLean, B. Swihart, L. Xiao, C. Crainiceanu, P. T. Reiss, and E. Cui. refund: Regression with Functional Data , Aug. 2010. URL https://CRAN.R-project.org/package=refund. Institution: Comprehensive R Archive Network

    work page 2010

  13. [13]

    Goldsmith, S

    J. Goldsmith, S. Greven, and C. Crainiceanu. Corrected confidence bands for functional data using principal components. Biometrics, 69 0 (1): 0 41--51, 2013

    work page 2013

  14. [14]

    Goldsmith, V

    J. Goldsmith, V. Zipunnikov, and J. Schrack. Generalized Multilevel Function -on- Scalar Regression and Principal Component Analysis . Biometrics, 71 0 (2): 0 344--353, June 2015. ISSN 0006-341X. doi:10.1111/biom.12278. URL https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4479975/

    work page doi:10.1111/biom.12278 2015

  15. [15]

    Golovkine, E

    S. Golovkine, E. Gunning, A. J. Simpkin, and N. Bargary. On the estimation of the number of components in multivariate functional principal component analysis. Communications in Statistics - Simulation and Computation, 0 0 (0): 0 1--9, Feb. 2025. ISSN 0361-0918. doi:10.1080/03610918.2025.2459862. URL https://doi.org/10.1080/03610918.2025.2459862. Publishe...

    work page doi:10.1080/03610918.2025.2459862 2025

  16. [16]

    G. H. Golub and C. F. V. Loan. Matrix computations (2nd edition), volume 74. Johns Hopkins Press, 1989. ISBN 0-8018-3772-3. doi:10.2307/3619868

    work page doi:10.2307/3619868 1989

  17. [17]

    and Morgan, K

    C. Happ and S. Greven. Multivariate Functional Principal Component Analysis for Data Observed on Different ( Dimensional ) Domains . Journal of the American Statistical Association, 113 0 (522): 0 649--659, Apr. 2018. ISSN 0162-1459. doi:10.1080/01621459.2016.1273115. URL https://doi.org/10.1080/01621459.2016.1273115. Publisher: ASA Website \_eprint: http...

    work page doi:10.1080/01621459.2016.1273115 2018

  18. [18]

    N. J. Higham and R. S. Schreiber. Fast polar decomposition of an arbitrary matrix. SIAM Journal on Scientific and Statistical Computing, 11 0 (4): 0 648--655, 1990. doi:10.1137/0911038. URL https://doi.org/10.1137/0911038

    work page doi:10.1137/0911038 1990

  19. [19]

    A. E. Ivanescu, C. M. Crainiceanu, and W. Checkley. Dynamic child growth prediction: A comparative methods approach. Statistical Modelling, 17 0 (6): 0 468--493, Dec. 2017. ISSN 1471-082X. doi:10.1177/1471082X17707619. URL https://doi.org/10.1177/1471082X17707619. Publisher: SAGE Publications India

    work page doi:10.1177/1471082x17707619 2017

  20. [20]

    A. E. Ivanescu, W. Checkley, and C. Crainiceanu. Outlier detection in dynamic functional models. 2024

    work page 2024

  21. [21]

    Jaganath, M

    D. Jaganath, M. Saito, R. H. Gilman, D. M. M. Queiroz, G. A. Rocha, V. Cama, L. Cabrera, D. Kelleher, H. J. Windle, J. E. Crabtree, and W. Checkley. First detected Helicobacter pylori infection in infancy modifies the association between diarrheal disease and childhood growth in Peru . Helicobacter, 19 0 (4): 0 272--279, Aug. 2014. ISSN 1523-5378. doi:10....

    work page doi:10.1111/hel.12130 2014

  22. [22]

    Jauch, P

    M. Jauch, P. D. Hoff, and D. B. Dunson. Monte Carlo Simulation on the Stiefel Manifold via Polar Expansion . Journal of Computational and Graphical Statistics, 30 0 (3): 0 622--631, Sept. 2021. ISSN 1061-8600. doi:10.1080/10618600.2020.1859382. URL https://doi.org/10.1080/10618600.2020.1859382. Publisher: Taylor & Francis \_eprint: https://doi.org/10.1080...

    work page doi:10.1080/10618600.2020.1859382 2021

  23. [23]

    Karhunen

    K. Karhunen. Uber lineare Methoden in der Wahrscheinlichkeitsrechnung . Annals of the Academy of Science Fennicae, 37 0 (Series A. I. Mathematics-Physics): 0 1--79, 1947

    work page 1947

  24. [24]

    Kimeldorf and G

    G. Kimeldorf and G. Wahba. A correspondence between bayesian estimation on stochastic processes and smoothing by splines. The Annals of Mathematical Statistics, 41 0 (2): 0 495--502, 1970

    work page 1970

  25. [25]

    D. Kosambi. Statistics in function space. Journal of the Indian Mathematical Society, 7: 0 77--88, 1943

    work page 1943

  26. [26]

    P. C. Lambert, A. J. Sutton, P. R. Burton, K. R. Abrams, and D. R. Jones. How vague is vague? A simulation study of the impact of the use of vague prior distributions in MCMC using WinBUGS . Statistics in Medicine, 24 0 (15): 0 2401--2428, Aug. 2005. ISSN 0277-6715. doi:10.1002/sim.2112

    work page doi:10.1002/sim.2112 2005

  27. [27]

    C. Li, L. Xiao, and S. Luo. Fast covariance estimation for multivariate sparse functional data. Stat (International Statistical Institute), 9 0 (1): 0 e245, Dec. 2020. ISSN 2049-1573. doi:10.1002/sta4.245. URL https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8276768/

    work page doi:10.1002/sta4.245 2020

  28. [28]

    X. Liu, H. Nassar, and K. PodgÓrski. Splinets -- efficient orthonormalization of the B -splines, Jan. 2020. URL http://arxiv.org/abs/1910.07341. arXiv:1910.07341 [math]

    work page arXiv 2020

  29. [29]

    M. Loève. Probability Theory , volume II of Graduate Texts in Mathematics . Springer-Verlag, 4th edition, 1978

    work page 1978

  30. [30]

    Y. Lu, X. Zhou, E. Cui, D. Rogers, C. M. Crainiceanu, J. Wrobel, and A. Leroux. Generalized Conditional Functional Principal Component Analysis , Nov. 2024. URL http://arxiv.org/abs/2411.10312. arXiv:2411.10312 [stat]

    work page arXiv 2024

  31. [31]

    T. H. Nolan, J. Goldsmith, and D. Ruppert. Bayesian Functional Principal Components Analysis via Variational Message Passing with Multilevel Extensions . Bayesian Analysis, -1 0 (-1): 0 1--27, Jan. 2023. ISSN 1936-0975, 1931-6690. doi:10.1214/23-BA1393. URL https://projecteuclid.org/journals/bayesian-analysis/advance-publication/Bayesian-Functional-Princi...

    work page doi:10.1214/23-ba1393 2023

  32. [32]

    T. H. Nolan, S. Richardson, and H. Ruffieux. Efficient Bayesian functional principal component analysis of irregularly-observed multivariate curves. Computational Statistics & Data Analysis, 203: 0 108094, Mar. 2025. ISSN 0167-9473. doi:10.1016/j.csda.2024.108094. URL https://www.sciencedirect.com/science/article/pii/S0167947324001786

    work page doi:10.1016/j.csda.2024.108094 2025

  33. [33]

    O’Sullivan

    F. O’Sullivan. A statistical perspective on ill-posed inverse problems (with discussion). Statistical Science, 1 0 (4): 0 505–527, 1986

    work page 1986

  34. [34]

    Papaspiliopoulos, G

    O. Papaspiliopoulos, G. O. Roberts, and M. Sköld. A General Framework for the Parametrization of Hierarchical Models . Statistical Science, 22 0 (1): 0 59--73, Feb. 2007. doi:https://doi.org/10.1214/088342307000000014. URL https://projecteuclid.org/journals/statistical-science/volume-22/issue-1/A-General-Framework-for-the-Parametrization-of-Hierarchical-M...

    work page doi:10.1214/088342307000000014 2007

  35. [35]

    Ramsay and B

    J. Ramsay and B. Silverman. Functional Data Analysis . Springer New York, NY, USA, 2005

    work page 2005

  36. [36]

    A. Redd. A comment on the orthogonalization of B -spline basis functions and their derivatives. Statistics and Computing, 22 0 (1): 0 251--257, Jan. 2012. ISSN 1573-1375. doi:10.1007/s11222-010-9221-0. URL https://doi.org/10.1007/s11222-010-9221-0

    work page doi:10.1007/s11222-010-9221-0 2012

  37. [37]

    D. Ruppert. Selecting the Number of Knots for Penalized Splines . Journal of Computational and Graphical Statistics, 11 0 (4): 0 735--757, Dec. 2002. ISSN 1061-8600. doi:10.1198/106186002853. URL https://doi.org/10.1198/106186002853

    work page doi:10.1198/106186002853 2002

  38. [38]

    Sartini, X

    J. Sartini, X. Zhou, L. Selvin, S. Zeger, and C. Crainiceanu. Fast Bayesian Functional Principal Components Analysis . Dec. 2024. doi:10.48550/arXiv.2412.11340. URL http://arxiv.org/abs/2412.11340. arXiv:2412.11340 [stat]

    work page doi:10.48550/arxiv.2412.11340 2024

  39. [39]

    Sharpe and N

    J. Sharpe and N. Fieller. Uncertainty in functional principal component analysis. Journal of Applied Statistics, 43 0 (12): 0 2295--2309, Sept. 2016. ISSN 0266-4763. doi:10.1080/02664763.2016.1140728. URL https://doi.org/10.1080/02664763.2016.1140728. Publisher: Taylor & Francis \_eprint: https://doi.org/10.1080/02664763.2016.1140728

    work page doi:10.1080/02664763.2016.1140728 2016

  40. [40]

    Staniswalis and J

    J. Staniswalis and J. Lee. Nonparametric regression analysis of longitudinal data. Journal of the American Statistical Association, 93 0 (444): 0 1403--1418, 1998

    work page 1998

  41. [41]

    S. D. Team. Stan modeling language users guide and reference manual 2.36, 2025. URL https://mc-stan.org/docs/reference-manual/transforms.html#ordered-vector

    work page 2025

  42. [42]

    Trendafilov and M

    N. Trendafilov and M. Gallo. Procrustes analysis (PA), pages 187--228. Springer International Publishing, Cham, 2021. ISBN 978-3-030-76974-1. doi:10.1007/978-3-030-76974-1_6. URL https://doi.org/10.1007/978-3-030-76974-1_6

    work page doi:10.1007/978-3-030-76974-1_6 2021

  43. [43]

    Confidence Intervals

    G. Wahba. Bayesian “Confidence Intervals” for the Cross-Validated Smoothing Spline . Journal of the Royal Statistical Society: Series B, 45 0 (1): 0 133--150, 1983

    work page 1983

  44. [44]

    G. A. Watson. The solution of orthogonal Procrustes problems for a family of orthogonally invariant norms. Advances in Computational Mathematics, 2 0 (4): 0 393--405, Sept. 1994. ISSN 1572-9044. doi:10.1007/BF02521606. URL https://doi.org/10.1007/BF02521606

    work page doi:10.1007/bf02521606 1994

  45. [45]

    L. Xiao, C. Li, W. Checkley, and C. Crainiceanu. Fast covariance estimation for sparse functional data. Statistics and Computing, 28 0 (3): 0 511--522, May 2018. ISSN 1573-1375. doi:10.1007/s11222-017-9744-8. URL https://doi.org/10.1007/s11222-017-9744-8

    work page doi:10.1007/s11222-017-9744-8 2018

  46. [46]

    Yao, H.-G

    F. Yao, H.-G. Müller, and J.-L. Wang. Functional Data Analysis for Sparse Longitudinal Data . Journal of the American Statistical Association, 100 0 (470): 0 577--590, June 2005. ISSN 0162-1459. doi:10.1198/016214504000001745. URL https://doi.org/10.1198/016214504000001745. Publisher: ASA Website \_eprint: https://doi.org/10.1198/016214504000001745

    work page doi:10.1198/016214504000001745 2005

  47. [47]

    J. Ye. Functional principal component models for sparse and irregularly spaced data by Bayesian inference. Journal of Applied Statistics, 51 0 (7): 0 1287--1317, May 2024. ISSN 0266-4763. doi:10.1080/02664763.2023.2197587. URL https://doi.org/10.1080/02664763.2023.2197587. Publisher: Taylor & Francis \_eprint: https://doi.org/10.1080/02664763.2023.2197587

    work page doi:10.1080/02664763.2023.2197587 2024

  48. [48]

    X. Zhou, E. Cui, J. Sartini, and C. Crainiceanu. Prediction Inference Using Generalized Functional Mixed Effects Models , Jan. 2025. URL http://arxiv.org/abs/2501.07842. arXiv:2501.07842 [stat]

    work page arXiv 2025