pith. machine review for the scientific record. sign in

arxiv: 2604.05453 · v1 · submitted 2026-04-07 · 🌌 astro-ph.HE

Recognition: 1 theorem link

· Lean Theorem

The NANOGrav 15 yr and 20 yr Datasets: Timing Events and Pulse Shape Changes

Ben Jacobson-Bell , James M. Cordes , Shami Chatterjee , Sashabaw Niedbalski , Gabriella Agazie , Akash Anumarlapudi , Anne M. Archibald , Zaven Arzoumanian
show 56 more authors
Jeremy G. Baier Paul T. Baker Paul R. Brook H. Thankful Cromartie Kathryn Crowter Megan E. DeCesar Paul B. Demorest Lankeswar Dey Timothy Dolch Elizabeth C. Ferrara William Fiore Emmanuel Fonseca Gabriel E. Freedman Nate Garver-Daniels Peter A. Gentile Joseph Glaser Deborah C. Good Jeffrey S. Hazboun Ross J. Jennings Megan L. Jones David L. Kaplan Matthew Kerr Michael T. Lam Bjorn Larsen Duncan R. Lorimer Georgia A. Lowes Jing Luo Ryan S. Lynch Ashley Martsen Alexander McEwen Maura A. McLaughlin Natasha McMann Bradley W. Meyers Patrick M. Meyers Cherry Ng Mason Ng David J. Nice Shania Nichols Daniel J. Oliver Timothy T. Pennucci Benetge B. P. Perera Nihan S. Pol Henri A. Radovan Scott M. Ransom Paul S. Ray Alexander Saffer Ann Schmiedekamp Carl Schmiedekamp Brent J. Shapiro-Albert Ingrid H. Stairs Kevin Stovall Abhimanyu Susobhanan Joseph K. Swiggum Mercedes S. Thompson Amir Tresnjic Haley M. Wahl
Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:57 UTC · model grok-4.3

classification 🌌 astro-ph.HE
keywords pulsar timingpulse shape variationsNANOGravprincipal component analysistiming eventsPSR J1713+0747pulse profiles
0
0 comments X

The pith

All three timing events in PSR J1713+0747 arise from changes in pulse shape.

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

The paper investigates whether discrete departures from the timing model in pulsar observations stem from alterations in the average pulse profile rather than external influences. By applying principal component analysis to pulse profiles from nine NANOGrav pulsars, including extended data for two, it recovers the three known events in PSR J1713+0747 as morphological shifts and confirms similar changes in other sources. It also ranks and discusses four new candidate events while recovering known slow variations and noting a recurrence in one pulsar. This connection matters because precise pulse timing underpins efforts to detect nanohertz gravitational waves, and unaccounted profile changes can mimic or mask signals.

Core claim

All three discrete timing events in PSR J1713+0747 correspond to morphological changes in pulse shape, recovered via PCA along with known events in other pulsars and four highly ranked candidate events.

What carries the argument

Principal component analysis of pulse profiles to isolate and rank morphological variations tied to timing residuals.

If this is right

  • Slow pulse shape variations are confirmed and tracked in PSR J1643-1224, PSR J1903+0327, and PSR B1937+21.
  • An unexpected recurrence of a slow variation occurs after roughly 10 years in PSR B1937+21.
  • Four additional highly ranked candidate events are identified for further study in the nine-pulsar sample.
  • Accounting for these changes can refine timing models used in pulsar timing arrays.

Where Pith is reading between the lines

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

  • Routine PCA monitoring of pulse profiles could become a standard step in processing pulsar timing data to reduce systematic errors.
  • If profile changes prove recurrent, they may follow patterns that allow predictive corrections across multiple pulsars.
  • Extending the analysis to the full 20-year dataset for more sources would test whether such events are common enough to affect gravitational-wave searches.

Load-bearing premise

The principal components extracted by PCA capture the actual pulse shape changes responsible for the timing events rather than noise or calibration effects.

What would settle it

Subtracting the timing offsets predicted by the top principal components from the observed residuals for PSR J1713+0747 and checking whether the three discrete events disappear while the overall timing precision improves.

Figures

Figures reproduced from arXiv: 2604.05453 by Abhimanyu Susobhanan, Akash Anumarlapudi, Alexander McEwen, Alexander Saffer, Amir Tresnjic, Anne M. Archibald, Ann Schmiedekamp, Ashley Martsen, Benetge B. P. Perera, Ben Jacobson-Bell, Bjorn Larsen, Bradley W. Meyers, Brent J. Shapiro-Albert, Carl Schmiedekamp, Cherry Ng, Daniel J. Oliver, David J. Nice, David L. Kaplan, Deborah C. Good, Duncan R. Lorimer, Elizabeth C. Ferrara, Emmanuel Fonseca, Gabriel E. Freedman, Gabriella Agazie, Georgia A. Lowes, Haley M. Wahl, Henri A. Radovan, H. Thankful Cromartie, Ingrid H. Stairs, James M. Cordes, Jeffrey S. Hazboun, Jeremy G. Baier, Jing Luo, Joseph Glaser, Joseph K. Swiggum, Kathryn Crowter, Kevin Stovall, Lankeswar Dey, Mason Ng, Matthew Kerr, Maura A. McLaughlin, Megan E. DeCesar, Megan L. Jones, Mercedes S. Thompson, Michael T. Lam, Natasha McMann, Nate Garver-Daniels, Nihan S. Pol, Patrick M. Meyers, Paul B. Demorest, Paul R. Brook, Paul S. Ray, Paul T. Baker, Peter A. Gentile, Ross J. Jennings, Ryan S. Lynch, Sashabaw Niedbalski, Scott M. Ransom, Shami Chatterjee, Shania Nichols, Timothy Dolch, Timothy T. Pennucci, William Fiore, Zaven Arzoumanian.

Figure 1
Figure 1. Figure 1: Time series of projection magnitudes (that is, dot products, Di) of PSR J1713+0747 pulse profile residuals (δU) along each of the first five PCs of the dataset. The dot products are normalized by the square root of each PC’s eigenvalue, σi, for comparison across PCs. The vertical gray lines give the epochs reported by P. B. Demorest et al. (2013), M. T. Lam et al. (2018), and H. Xu et al. (2021) for their … view at source ↗
Figure 2
Figure 2. Figure 2: Left panel: the mean pulse profile for PSR J1713+0747 in our dataset, followed by the first ten PCs. Elements are zeroed outside the central 50% of pulse phase. Successive PCs capture decreasing amounts of the variance in the dataset, with the result that after the first several PCs, they begin to resemble white noise. Right panel: the square roots of the eigenvalues corresponding to each of the first ten … view at source ↗
Figure 3
Figure 3. Figure 3: Scatter plot of the dot products, Di, of PSR J1713+0747 pulse profile residuals (δU) with their first two PCs, scaled by σi. For clarity, only the Di of 1400 MHz profiles are shown. Color denotes the epoch of observation, beginning 200 d before the 2021 event. A discontinuity is visible around MJD 59320; the points then trace a curve back toward the locus of pre-event points, but do not reach it. In the al… view at source ↗
Figure 4
Figure 4. Figure 4: Toy model of our matched filtering framework on two basic time series–like vectors of length 2048. The top panel shows an evenly sampled (δt = 30) FRED event with decay time τ = 300; the bottom panel shows a single-pixel spike (a delta function) with the same sampling. Both time series are convolved with a FRED template and a DERF template, both with τ = 300, and the CCFs overplotted in the corresponding p… view at source ↗
Figure 5
Figure 5. Figure 5: Top panel: Application of our matched filtering framework to the D2 time series for PSR J1713+0747 in the 1400 MHz subband. For demonstration, we use τ = 100 d for our template decay time. Middle panel: Best-fit decay￾ing exponentials to an interval of D2 following the epoch of each event. The interval is 1 yr for the first two events and ∼1.7 yr for the third, since the event is clearly still occurring by… view at source ↗
Figure 6
Figure 6. Figure 6: Top panel: CCF responses of PC 2 projection magnitudes of PSR J1713+0747 data with FRED templates with τ ranging from 50 to 300 d in increments of 10 d. Bottom panel: A heatmap of CCF responses over frequency with τ0 fixed at 100 d. The epochs of known events are given by the red ticks and define the zero points along the top x-axis. with decay time τ0. The expected response for a true FRED event is ℓ = 1.… view at source ↗
Figure 7
Figure 7. Figure 7: Time series of ℓ and α values calculated for pairs of maxima or minima between the FRED CCF and the DERF CCF for each of the first six PCs for each pulsar in our sample. Circles are pairs of maxima; squares are pairs of minima. The size of each marker gives the S/N of the FRED CCF extremum for that pair, though these S/N values are not normalized by eigenvalues on a per-PC basis as the panels in [PITH_FUL… view at source ↗
Figure 8
Figure 8. Figure 8: Distribution of ζmin for all candidate events in each pulsar in our sample (see Tables 2 and 3). The four known events are indicated with unique markers. Lower ζmin values indicate better candidates, with events to the left of the gray shaded region considered significant. Any time-symmetric pulse shape change event will likewise have α/(e/2) ≈ 0.736, but small asymmetries may elevate α into our search ran… view at source ↗
Figure 9
Figure 9. Figure 9: The four known pulse shape change events as seen in our PCA time series. Note the factor of 10 difference in y-axis scale for the bottom-right panel, showing the 2021 event in PSR J1713+0747, versus all other panels; despite this factor of 10, the ζ scores for the 2016 and 2021 events in PSR J1713+0747 are almost the same because ζ has no direct S/N dependence. In all panels, the scatter plots are snippets… view at source ↗
Figure 10
Figure 10. Figure 10: The four pulse shape change event candidates found by our automated search with ζ < 0.25, the threshold set by the 2008 event in PSR J1713+0747. FRED shapes are overlaid in each case with the best-fit amplitude and decay time for each 100 MHz frequency channel in which the candidate was detected. Like in [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Spectral dependence of τ and S/N for each event isolated by the 1D mean-shift clustering algorithm in PC 2 of PSR J1713+0747 data. For greater y-axis resolution, the τ and S/N shown here are the result of FRED template fitting to the Di time series around the detection epoch. Vertical error bars indicate the 1σ error in this fit, while horizontal error bars indicate the subband width in frequency. Power l… view at source ↗
Figure 13
Figure 13. Figure 13: Top panel: Time series of projection magnitudes of PSR B1937+21 pulse profile residuals along their third PC. As in [PITH_FULL_IMAGE:figures/full_fig_p018_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Scatter plots showing the correlations between PC magnitudes (dot products) for PSR B1937+21 pulse profiles recorded with the GBT and Arecibo at 1400 MHz between MJDs 55600 and 55850 (the orange band of [PITH_FULL_IMAGE:figures/full_fig_p018_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Left: Angular separation of each pulsar in our sample from the Sun over 1 yr. Right: Range of angular separations for each pulsar. The coordinates used for each pulsar are those given in their names (e.g., α = 17h 13m00s and δ = 07◦ 47′ 00′′ for PSR J1713+0747) and are inaccu￾rate by up to a few degrees (for PSR B1937+21). The solar ephemeris is due to JPL Horizons. 0.1 1 period [yr] 0.00 0.05 0.10 0.15 p… view at source ↗
Figure 16
Figure 16. Figure 16: Lomb–Scargle periodograms of the 800 MHz dot products for each of the first five PCs of PSR B1937+21 (in red) and the average of the five (in black). The peak on the right-hand side is at ∼1.6 yr. is not unexpected given the low magnitude of the pulse shape variation above 1 GHz. On the basis of these correlations, combined with the fact that comparable variation does not seem to occur at 820 MHz for any … view at source ↗
Figure 17
Figure 17. Figure 17: Spectral dependence of the (top panel) 2011 and (bottom panel) 2021 pulse shape change events in each of the first five PCs of PSR B1937+21. The y-axes give the ampli￾tude of each event at each frequency, calculated by fitting a parabola to the points at that frequency in the respective time range. Vertical error bars are ±1σ; horizontal error bars are the subband width in frequency. The subscript e in th… view at source ↗
Figure 18
Figure 18. Figure 18: Effect of nonuniform time sampling on FRED–DERF amplitude ratio α, dimensionless lag ℓ, and ζ score (Equation 12). Here we show 500 trials of the same FRED event sampled at different average rates ⟨δtsamp⟩, binned according to various prescriptions, then matched with a FRED and a DERF template of the same decay time τ . Each sample was allowed to occur earlier or later than the fiducial time by up to ⟨δts… view at source ↗
read the original abstract

The average pulse shape of a pulsar is typically stable over decadal timescales, enabling estimation of pulse times of arrival to better than a small fraction of the pulse width using matched filtering techniques. However, in North American Nanohertz Observatory for Gravitational Waves (NANOGrav) observations of PSR J1713+0747, three discrete timing events that depart from the prevailing timing model have been seen in the last 20 yr. All three correspond to morphological changes in pulse shape. Using principal component analysis, we analyze the pulse profiles of nine NANOGrav pulsars, including seven with profiles from the 15 yr dataset and two with additional profiles from the forthcoming 20 yr dataset. We recover the three known pulse shape change events in PSR J1713+0747 and another previously known event in PSR J1643$-$1224. We implement a ranking metric for candidate events and address four highly ranked candidates in this nine-pulsar sample. We also recover known slow pulse shape variations in PSR J1643$-$1224, PSR J1903+0327, and PSR B1937+21 and report an unexpected recurrence after ~10 yr of one such variation in PSR B1937+21.

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. This manuscript applies principal component analysis (PCA) to pulse profiles from nine NANOGrav pulsars using the 15 yr and 20 yr datasets. It recovers three previously reported discrete timing events in PSR J1713+0747 (all linked to pulse shape changes) and one known event in PSR J1643-1224, implements a ranking metric to identify four additional candidate events, and reports recovery of known slow profile variations in several pulsars plus an unexpected recurrence after ~10 yr in PSR B1937+21.

Significance. Recovery of multiple known events provides internal validation for the PCA approach on public timing data. If the new candidates and recurrence hold under further scrutiny, the work would strengthen understanding of profile variability as a source of timing noise, with direct relevance to improving pulsar timing array sensitivity for nanohertz gravitational waves. The observational nature and use of standard PCA are strengths.

major comments (3)
  1. [PCA analysis and ranking metric sections] The central claim that timing events arise from morphological changes recovered as leading principal components requires that PCA isolates astrophysical profile variations rather than epoch-dependent calibration residuals or noise. The manuscript should include explicit null tests (e.g., phase-randomized profiles or calibration-only simulations) to quantify the false-positive rate of the ranking metric, particularly for the four new candidates; without them, alignment between timing residuals and PC amplitudes could reflect shared systematics.
  2. [Methods and results for candidate ranking] The ranking metric for candidate selection and the number of principal components retained are free parameters. The paper should report sensitivity tests showing how the identification of the four highly ranked candidates and the recurrence in PSR B1937+21 changes with variations in these choices, including quantitative thresholds and error bars on event significance.
  3. [Results for J1713+0747 and J1643-1224] For the recovered events in J1713+0747 and J1643-1224, the manuscript should provide explicit statistical measures (e.g., correlation coefficients between PC amplitudes and timing event epochs or p-values) demonstrating that the correspondence exceeds what would be expected from random alignment, beyond the qualitative recovery statement.
minor comments (2)
  1. [Abstract] The abstract would benefit from brief mention of the statistical thresholds or controls used in the ranking metric to allow readers to assess the strength of the new candidate claims.
  2. [Figures] Figure captions should clarify the exact definition of the ranking metric and any normalization applied to PC amplitudes for reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments on our manuscript. We address each major comment in turn below and will revise the manuscript to incorporate the suggested improvements.

read point-by-point responses

  1. Referee: [PCA analysis and ranking metric sections] The central claim that timing events arise from morphological changes recovered as leading principal components requires that PCA isolates astrophysical profile variations rather than epoch-dependent calibration residuals or noise. The manuscript should include explicit null tests (e.g., phase-randomized profiles or calibration-only simulations) to quantify the false-positive rate of the ranking metric, particularly for the four new candidates; without them, alignment between timing residuals and PC amplitudes could reflect shared systematics.

    Authors: We agree that explicit null tests are needed to demonstrate that the leading principal components and the ranking metric primarily capture astrophysical profile variations rather than shared systematics. Although the successful recovery of multiple previously reported events provides supporting evidence for the approach, we will add null tests using phase-randomized profiles and calibration-only simulations. These will be used to estimate the false-positive rate of the ranking metric, with particular attention to the four new candidate events. The revised manuscript will include these results. revision: yes

  2. Referee: [Methods and results for candidate ranking] The ranking metric for candidate selection and the number of principal components retained are free parameters. The paper should report sensitivity tests showing how the identification of the four highly ranked candidates and the recurrence in PSR B1937+21 changes with variations in these choices, including quantitative thresholds and error bars on event significance.

    Authors: We acknowledge that both the number of retained principal components and the details of the ranking metric are choices that require justification. In the revision we will conduct and report sensitivity tests that vary the number of principal components and the quantitative thresholds used in the ranking metric. These tests will show the stability of the four highly ranked candidates and the reported recurrence in PSR B1937+21, and will include error bars or confidence intervals on the significance of each event. revision: yes

  3. Referee: [Results for J1713+0747 and J1643-1224] For the recovered events in J1713+0747 and J1643-1224, the manuscript should provide explicit statistical measures (e.g., correlation coefficients between PC amplitudes and timing event epochs or p-values) demonstrating that the correspondence exceeds what would be expected from random alignment, beyond the qualitative recovery statement.

    Authors: The current manuscript presents the recovery of the known events in J1713+0747 and J1643-1224 as qualitative validation of the PCA method. We agree that quantitative statistical measures would make the correspondence more rigorous. In the revised version we will compute and report correlation coefficients between the relevant PC amplitudes and the timing residuals (or event epochs), together with p-values that quantify the probability of obtaining such alignments by chance. revision: yes

Circularity Check

0 steps flagged

No significant circularity in observational PCA analysis of pulsar profiles

full rationale

The paper applies standard principal component analysis to observational pulse profile data from NANOGrav timing datasets to recover known morphological changes and rank candidate events. No mathematical derivation, first-principles prediction, or parameter fitting is presented that reduces by construction to the input data or self-referential definitions. Known events are recovered empirically, and the ranking metric operates on PCA outputs without equating predictions to fitted inputs. The analysis is self-contained against external benchmarks of documented pulsar events and does not rely on load-bearing self-citations or imported uniqueness theorems.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The analysis rests on standard statistical decomposition of pulse profiles and domain assumptions about the correspondence between shape and timing; no new physical entities are introduced and free parameters are limited to choices in the PCA and ranking procedure.

free parameters (2)
  • Number of principal components retained
    Selected to capture dominant morphological variations; exact count not stated in abstract.
  • Ranking metric thresholds for candidate selection
    Used to identify the four highly ranked events; specific values or derivation not provided.
axioms (2)
  • domain assumption Pulse profile variations can be adequately represented by linear principal components
    Standard assumption when applying PCA to time-series pulse data.
  • domain assumption Discrete timing events arise from or coincide with pulse shape morphology changes
    Invoked to interpret the recovered events in J1713+0747 and other pulsars.

pith-pipeline@v0.9.0 · 5856 in / 1289 out tokens · 87718 ms · 2026-05-10T19:57:48.791345+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

46 extracted references · 46 canonical work pages

  1. [1]

    F., Anumarlapudi, A., et al

    Agazie, G., Alam, M. F., Anumarlapudi, A., et al. 2023, ApJL, 951, L9, doi: 10.3847/2041-8213/acda9a

    work page doi:10.3847/2041-8213/acda9a 2023

  2. [2]

    M., et al

    Agazie, G., Anumarlapudi, A., Archibald, A. M., et al. 2025, ApJ, 978, 168, doi: 10.3847/1538-4357/ad93aa

    work page doi:10.3847/1538-4357/ad93aa 2025

  3. [3]

    F., Arzoumanian, Z., Baker, P

    Alam, M. F., Arzoumanian, Z., Baker, P. T., et al. 2021, ApJS, 252, 4, doi: 10.3847/1538-4365/abc6a0

    work page doi:10.3847/1538-4365/abc6a0 2021

  4. [4]

    Goss, W. M. 1982, Nature, 300, 615, doi: 10.1038/300615a0

    work page doi:10.1038/300615a0 1982

  5. [5]

    Bartel, N., Morris, D., Sieber, W., & Hankins, T. H. 1982, ApJ, 258, 776, doi: 10.1086/160125

    work page doi:10.1086/160125 1982

  6. [6]

    R., Karastergiou, A., McLaughlin, M

    Brook, P. R., Karastergiou, A., McLaughlin, M. A., et al. 2018, ApJ, 868, 122, doi: 10.3847/1538-4357/aae9e3

    work page doi:10.3847/1538-4357/aae9e3 2018

  7. [7]

    2020, in American Astronomical Society Meeting Abstracts, Vol

    Bulatek, A., & White, S. 2020, in American Astronomical Society Meeting Abstracts, Vol. 235, American Astronomical Society Meeting Abstracts #235, 175.17

    work page 2020

  8. [8]

    2002, IEEE Transactions on Pattern Analysis and Machine Intelligence, 24, 603, doi: 10.1109/34.1000236

    Comaniciu, D., & Meer, P. 2002, IEEE Transactions on Pattern Analysis and Machine Intelligence, 24, 603, doi: 10.1109/34.1000236

    work page doi:10.1109/34.1000236 2002

  9. [9]

    Cordes, J. M. 2013, ApJ, 775, 47, doi: 10.1088/0004-637X/775/1/47

    work page doi:10.1088/0004-637x/775/1/47 2013

  10. [10]

    M., & Lazio, T

    Cordes, J. M., & Lazio, T. J. 1991, ApJ, 376, 123, doi: 10.1086/170261

    work page doi:10.1086/170261 1991

  11. [11]

    Demorest, P. B. 2007, Phd thesis, Univ. California, Berkeley

    work page 2007

  12. [12]

    B., Ferdman, R

    Demorest, P. B., Ferdman, R. D., Gonzalez, M. E., et al. 2013, ApJ, 762, 94, doi: 10.1088/0004-637X/762/2/94

    work page doi:10.1088/0004-637x/762/2/94 2013

  13. [13]

    1979, ApJ, 234, 1100, doi: 10.1086/157593

    Detweiler, S. 1979, ApJ, 234, 1100, doi: 10.1086/157593

    work page doi:10.1086/157593 1979

  14. [14]

    A., Wahl, H

    Dey, L., McLaughlin, M. A., Wahl, H. M., et al. 2024, ApJ, 977, 114, doi: 10.3847/1538-4357/ad8c24

    work page doi:10.3847/1538-4357/ad8c24 2024

  15. [15]

    2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol

    DuPlain, R., Ransom, S., Demorest, P., et al. 2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7019, Advanced Software and Control for Astronomy II, ed. A. Bridger & N. M. Radziwill, 70191D, doi: 10.1117/12.790003

    work page doi:10.1117/12.790003 2008

  16. [16]

    A., Agazie, G., et al

    Fiore, W., McLaughlin, M. A., Agazie, G., et al. 2025, ApJ, 984, 139, doi: 10.3847/1538-4357/adc255

    work page doi:10.3847/1538-4357/adc255 2025

  17. [17]

    M., Lam, M

    Geiger, A., Cordes, J. M., Lam, M. T., et al. 2025, ApJ, 986, 191, doi: 10.3847/1538-4357/add0b6

    work page doi:10.3847/1538-4357/add0b6 2025

  18. [18]

    A., McLaughlin, M

    Gentile, P. A., McLaughlin, M. A., Demorest, P. B., et al. 2018, ApJ, 862, 47, doi: 10.3847/1538-4357/aac9c9

    work page doi:10.3847/1538-4357/aac9c9 2018

  19. [19]

    J., Shannon, R

    Goncharov, B., Reardon, D. J., Shannon, R. M., et al. 2021, MNRAS, 502, 478, doi: 10.1093/mnras/staa3411

    work page doi:10.1093/mnras/staa3411 2021

  20. [20]

    2001, PASP, 113, 1274, doi: 10.1086/323289

    Heiles, C., Perillat, P., Nolan, M., et al. 2001, PASP, 113, 1274, doi: 10.1086/323289

    work page doi:10.1086/323289 2001

  21. [21]

    W., & Downs, G

    Hellings, R. W., & Downs, G. S. 1983, ApJL, 265, L39, doi: 10.1086/183954

    work page doi:10.1086/183954 1983

  22. [22]

    J., Cordes, J

    Jennings, R. J., Cordes, J. M., & Chatterjee, S. 2024a, arXiv e-prints, arXiv:2411.00236, doi: 10.48550/arXiv.2411.00236

    work page doi:10.48550/arxiv.2411.00236

  23. [23]

    J., Cordes, J

    Jennings, R. J., Cordes, J. M., Chatterjee, S., et al. 2024b, ApJ, 964, 179, doi: 10.3847/1538-4357/ad2930

    work page doi:10.3847/1538-4357/ad2930

  24. [24]

    J., Spiewak, R., Lyne, A

    Keith, M. J., Spiewak, R., Lyne, A. G., et al. 2025, MNRAS, 540, 2486, doi: 10.1093/mnras/staf886

    work page doi:10.1093/mnras/staf886 2025

  25. [25]

    Lorimer, D. R. 2006, Science, 312, 549, doi: 10.1126/science.1124060

    work page doi:10.1126/science.1124060 2006

  26. [26]

    Lam, M. T. 2017,, Astrophysics Source Code Library, record ascl:1706.011

    work page 2017

  27. [27]

    T., Ellis, J

    Lam, M. T., Ellis, J. A., Grillo, G., et al. 2018, ApJ, 861, 132, doi: 10.3847/1538-4357/aac770

    work page doi:10.3847/1538-4357/aac770 2018

  28. [28]

    X., Lin, H.-H., Luo, J., et al

    Lin, F. X., Lin, H.-H., Luo, J., et al. 2021, MNRAS, 508, 1115, doi: 10.1093/mnras/stab2529

    work page doi:10.1093/mnras/stab2529 2021

  29. [29]

    Lin, H.-H., Masui, K., Pen, U.-L., & Peterson, J. B. 2018, MNRAS, 475, 1323, doi: 10.1093/mnras/stx3239

    work page doi:10.1093/mnras/stx3239 2018

  30. [30]

    Lomb, N. R. 1976, Ap&SS, 39, 447, doi: 10.1007/BF00648343

    work page doi:10.1007/bf00648343 1976

  31. [31]

    2023, in American Astronomical Society Meeting Abstracts, Vol

    Lynch, R., White, S., Simon, B., et al. 2023, in American Astronomical Society Meeting Abstracts, Vol. 241, American Astronomical Society Meeting Abstracts #241, 101.04

    work page 2023

  32. [32]

    2010, Science, 329, 408, doi: 10.1126/science.1186683 27

    Lyne, A., Hobbs, G., Kramer, M., Stairs, I., & Stappers, B. 2010, Science, 329, 408, doi: 10.1126/science.1186683 27

    work page doi:10.1126/science.1186683 2010

  33. [33]

    2003, ApJ, 582, 972, doi: 10.1086/344816

    Maitia, V., Lestrade, J.-F., & Cognard, I. 2003, ApJ, 582, 972, doi: 10.1086/344816

    work page doi:10.1086/344816 2003

  34. [34]

    2026 in prep

    Martsen, A., et al. 2026 in prep

    work page 2026

  35. [35]

    2024, ApJ, 974, 10, doi: 10.3847/1538-4357/ad6a51

    Ocker, S. K., Anderson, L. D., Lazio, T. J. W., Cordes, J. M., & Ravi, V. 2024, ApJ, 974, 10, doi: 10.3847/1538-4357/ad6a51

    work page doi:10.3847/1538-4357/ad6a51 2024

  36. [36]

    2011, Journal of Machine Learning Research, 12, 2825

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825

    work page 2011

  37. [37]

    Pennucci, T. T. 2019, ApJ, 871, 34, doi: 10.3847/1538-4357/aaf6ef

    work page doi:10.3847/1538-4357/aaf6ef 2019

  38. [38]

    T., Demorest, P

    Pennucci, T. T., Demorest, P. B., & Ransom, S. M. 2016,, Astrophysics Source Code Library, record ascl:1606.013 http://ascl.net/1606.013

    work page 2016

  39. [39]

    Sazhin, M. V. 1978, Soviet Ast., 22, 36

    work page 1978

  40. [40]

    Scargle, J. D. 1982, ApJ, 263, 835, doi: 10.1086/160554

    work page doi:10.1086/160554 1982

  41. [41]

    M., Cordes, J

    Shannon, R. M., Cordes, J. M., Metcalfe, T. S., et al. 2013, ApJ, 766, 5, doi: 10.1088/0004-637X/766/1/5

    work page doi:10.1088/0004-637x/766/1/5 2013

  42. [42]

    M., Lentati, L

    Shannon, R. M., Lentati, L. T., Kerr, M., et al. 2016, ApJL, 828, L1, doi: 10.3847/2041-8205/828/1/L1

    work page doi:10.3847/2041-8205/828/1/l1 2016

  43. [43]

    Taylor, J. H. 1992, Philosophical Transactions of the Royal Society of London Series A, 341, 117, doi: 10.1098/rsta.1992.0088 The NANOGrav Collaboration. 2025,, 2.1.0 Zenodo, doi: 10.5281/zenodo.16051178 van Straten, W. 2004, ApJS, 152, 129, doi: 10.1086/383187

    work page doi:10.1098/rsta.1992.0088 1992

  44. [44]

    M., McLaughlin, M

    Wahl, H. M., McLaughlin, M. A., Gentile, P. A., et al. 2022, ApJ, 926, 168, doi: 10.3847/1538-4357/ac4045

    work page doi:10.3847/1538-4357/ac4045 2022

  45. [45]

    X., Burgay, M., et al

    Xu, H., Huang, Y. X., Burgay, M., et al. 2021, The Astronomer’s Telegram, 14642, 1

    work page 2021

  46. [46]

    X., & Bloom, J

    You, X. P., Coles, W. A., Hobbs, G. B., & Manchester, R. N. 2012, MNRAS, 422, 1160, doi: 10.1111/j.1365-2966.2012.20688.x

    work page doi:10.1111/j.1365-2966.2012.20688.x 2012