Pith. sign in

REVIEW 3 major objections 5 minor 63 references

Fitting a spatiotemporal Hawkes process to 412,376 U.S. gun-violence events from 2014–2024, this paper estimates a branching number of 1.15 and concludes that gun violence is explosive: each event births an average of 1.15 additional events

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

T0 review · deepseek-v4-flash

2026-08-01 21:23 UTC pith:UVBZ6DEI

load-bearing objection A useful scaling demo undermined by a non-causal background term that invalidates the explosive branching ratio claim. the 3 major comments →

arxiv 2607.16081 v1 pith:UVBZ6DEI submitted 2026-07-17 stat.CO stat.APstat.ME

Scaling Hawkes Processes

classification stat.CO stat.APstat.ME MSC 60G5562M3062F15
keywords Hawkes processspatiotemporal point processself-excitationbranching numbergun violenceBayesian inferencehigh-performance computingcontagion
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Using a spatiotemporal Hawkes process—a point-process model in which past events raise the future event rate—this paper argues that U.S. gun violence between 2014 and 2024 is explosive. Fitting the model to 412,376 recorded incidents with a multi-GPU Bayesian sampler, the authors estimate a branching number of 1.15 (95% credible interval 1.15–1.16), meaning each act of violence is expected to produce about 1.15 further acts. A branching number above 1 is the threshold for a self-sustaining contagion: even with no change in background conditions, event counts can grow through triggered chains. The paper's broader point is computational: with high-performance parallelization, likelihood-based Bayesian Hawkes inference that normally scales quadratically can be applied to big public-health data, yielding uncertainty-quantified estimates of contagion parameters.

Core claim

On its own terms, the paper's central discovery is that a self-exciting model of American gun violence has a branching number greater than one. The authors specify a spatiotemporal Hawkes intensity whose triggering kernel is an exponential-in-time, Gaussian-in-space density, so the kernel's total integral is 1 and the self-excitatory weight ξ0 equals the branching number β—the expected number of offspring per event. After Bayesian inference on 412,376 events, the posterior median of ξ0 is 1.15 with a 95% credible interval (1.15, 1.16), leading the authors to conclude that 'each event births an additional 1.15 events.' They also estimate that triggered children occur at a median spatial dista

What carries the argument

The central object is the branching number β, realized here as the self-excitatory weight ξ0 in the conditional intensity (20). The intensity's background term is a time-smoothed spatial density over observed events, and its triggering term is a product of an exponential temporal kernel and a spherical Gaussian spatial kernel; because each triggering kernel integrates to 1, ξ0 is exactly the expected number of offspring an event produces. The inference machinery is a multi-GPU, many-chain adaptive Metropolis-Hastings algorithm that parallelizes the O(N^2) likelihood evaluation, making Bayesian posterior sampling on 412,376 events feasible. The parameter's posterior distribution—not a point e

Load-bearing premise

The load-bearing premise is that equation (20) is a genuine conditional intensity, meaning the rate at any moment depends only on events that have already happened; as written its background term sums over all 412,376 events including future ones, so this premise is not satisfied and, if left unrepaired, the likelihood and the branching-number interpretation both collapse.

What would settle it

Re-fit the model with the background term restricted to t_n < t, making the intensity strictly causal, and recompute the posterior of ξ0; if the 95% credible interval then includes or falls below 1, the explosive conclusion is refuted. A complementary check is to simulate from the fitted model and compare the distribution of future event counts with holdout data.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the branching number is 1.15, gun violence is supercritical: in expectation each generation is 15% larger than the last, so triggered chains alone can sustain and amplify violence without new external drivers.
  • The estimated temporal scale of about 1.5 years and spatial scale of roughly a mile mean contagion operates through slow, local chains, so short-term or jurisdiction-wide analyses may miss most of the triggered burden.
  • Because the triggering kernels are normalized densities, the same fitted model can be used to simulate counterfactual reductions in ξ0 and quantify how much future violence is attributable to self-excitation.
  • The paper explicitly cautions that the inference is provisional, since no competing models were compared and the extremely narrow credible intervals are 'particularly distressing'; it argues that forecasting accuracy should be the decisive test.

Where Pith is reading between the lines

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

  • One editorial caution: equation (20)'s background term sums over all events with only an inequality t≠t_n, not t_n<t, so the intensity at time t depends on future events; as written this is not a causal, adapted Hawkes intensity, and the branching-number interpretation depends on repairing that.
  • If the explosive estimate survives a causal correction and model comparison, a natural extension is to estimate intervention effects by simulating the fitted process with reduced ξ0—for example, halving the triggering weight—and measuring the drop in long-run event counts; this would turn the branching number into a policy lever.
  • The same scaling recipe could be applied to other self-exciting social phenomena with large spatiotemporal datasets, such as overdoses, protests, or retaliatory crime, where the branching number would provide a comparable contagion measure.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper reviews computational strategies for fitting Hawkes processes (methods of moments, ML, EM, MCMC, SGD) with emphasis on O(N^2) bottlenecks, and then applies a multi-GPU Bayesian inference framework to 412,376 U.S. gun violence events (2014–2024) using a spatiotemporal Hawkes process with conditional intensity (20). The central empirical claim is that the process is 'explosive': the posterior median of the self-excitatory weight is ξ0 = 1.15 (95% CrI 1.15–1.16), which the authors equate with the branching number because the triggering kernels are normalized densities. The paper also reports computational scaling results (31 adaptive MCMC chains, ~10 hours on 32 GPUs, Rhat < 1.003, high ESS).

Significance. If valid, the computational demonstration is genuinely useful: fitting a nonlinear likelihood to 412k events with good MCMC diagnostics in 10 hours is a nontrivial scalability result, and the review of fitting strategies is pedagogically clear. However, the scientific conclusion of supercriticality rests on a model whose conditional intensity is not adapted to the event history. The technical flaw is not cosmetic: it invalidates the Hawkes likelihood and the immigrant–offspring branching interpretation. The paper also contains an explicit self-acknowledged limitation ('we cannot fully trust our inference') in the Discussion. The computational contribution is worth preserving, but the explosive-violence conclusion is unsupported as stated.

major comments (3)
  1. [§3, Eq. (20)] The background term in Eq. (20) is λ0/|X| ∑_{n=1}^N I[t≠t_n] φ1(t|t_n, τ_t), with no restriction t_n < t. Thus the intensity at time t uses event times strictly after t. This violates the definition of a conditional intensity in Eq. (1) and the causal structure required for the Hawkes–Oakes cluster representation. Consequently, the likelihood (11) is not the likelihood of any Hawkes process, and the posterior distribution of ξ0 cannot be interpreted as a branching number. The paper neither flags nor justifies this non-adaptivity; this is load-bearing for the claim that the gun violence process is explosive.
  2. [§3, Table 1 and surrounding text] The identification 'ξ0 corresponds exactly to the branching number β' relies on Eq. (2) and on the triggering kernel being a normalized density. That identification is only valid for a causal Hawkes process with an immigrant–offspring representation. Because Eq. (20) is non-adapted, this equality is not established. Even after correcting the background term to a causal form, the authors should verify that the branching-number interpretation holds for the specific spatiotemporal domain (e.g., the integral of the spatial kernel over the bounded spatial domain X) before concluding supercriticality.
  3. [§4, Discussion] The authors state that they 'cannot fully trust our inference, not having compared our model to other competitors' and that the narrow credible intervals in Table 1 are 'particularly distressing.' These are not minor caveats: the paper's headline conclusion (branching number 1.15, explosive process) is exactly the kind of causal claim that demands model comparison or predictive validation. The current manuscript presents this as a firm conclusion in Section 3 and the abstract, while the Discussion explicitly undermines it. At minimum, the abstract and conclusion should be softened to match the stated uncertainty, or the analysis should be supplemented with the forecasting comparison the authors themselves propose.
minor comments (5)
  1. [§2, Example 2] Typo: 'condidtional intensity' should be 'conditional intensity' before Eq. (8).
  2. [§2, Example 4] The LSTM update equations contain duplicated lines ('in+1 ← ...' appears twice) and inconsistent subscripts; this makes the parameterization harder to follow.
  3. [§2, Example 5] Typo: 'emprical' should be 'empirical' in the method-of-moments discussion.
  4. [§2, Example 5] Equation (19) writes 'bΛ' for the Monte Carlo estimate, but the text refers to 'bΛ' and '∇bΛ' interchangeably; clarify the notation.
  5. [§3, Figure 2] The right panel of Figure 2 is described as using 412,376 data points, but no details are given about how the speedup was measured (e.g., number of replicates, burn-in, hardware variability). A sentence on measurement methodology would help.

Circularity Check

0 steps flagged

No significant circularity: the branching-number estimate is a fitted parameter with a definitional interpretation, not a prediction derived from its own inputs.

full rationale

The paper's central claim about explosiveness is an interpretation of the posterior estimate of ξ0 in Eq. (20). The statement that 'Because the temporal and spatial kernels used within the triggering kernel are normalized densities, the self-excitatory weight ξ0 corresponds exactly to the branching number β of (2)' is a reparameterization, not a circular derivation: ξ0 is estimated from the data, and its interpretation as the branching number follows from the model specification. This is an empirical finding about a fitted parameter, not a 'prediction' that is forced by construction. Self-citations to Ko et al. (2024) and Holbrook et al. (2021) supply computational machinery (multi-GPU likelihood evaluation and adaptive MH) rather than evidence for the scientific conclusion, so they are not load-bearing in a circular sense. The concern raised about Eq. (20)'s background term including future events (no restriction t_n < t) is a serious model-validity problem: the fitted object may not be a causal Hawkes process and the likelihood may be misspecified. But that is a correctness risk, not a circularity: the conclusion does not reduce to the model's inputs by construction. The paper itself flags important limitations, including lack of model comparison and 'extremely narrow posterior credible intervals,' which further supports treating the result as an uncertain empirical estimate rather than a circular derivation. No circular step meeting the evidentiary standard is present.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 0 invented entities

The central empirical result requires five fitted parameters and several domain assumptions, but the paper introduces no new entities. The branching ratio is a fitted parameter, not an independent prediction, so the scientific content is an application of a known model rather than a derivation.

free parameters (6)
  • λ0 (background weight) = 0.0231 (posterior median)
    Fitted background rate component; posterior median reported in Table 1.
  • τ_t (background temporal lengthscale) = 2.62 weeks
    Fitted temporal smoothing scale for the background term in Eq. 20; Table 1.
  • ξ0 (self-excitatory weight / branching number) = 1.15
    Fitted self-excitation weight; equals branching number because triggering kernels are normalized densities; Table 1.
  • ω_x (triggering spatial lengthscale) = 0.0205 degrees
    Fitted spatial offspring distance scale; Table 1.
  • ω_t (triggering temporal lengthscale) = 75.8 weeks
    Fitted temporal offspring delay scale; Table 1.
  • Prior hyperparameters from Ko et al. (2024) = not stated
    Chosen by hand from an earlier paper; authors say priors have little influence in big data, but exact values are needed for replication.
axioms (5)
  • standard math Hawkes/Oakes immigrant-offspring representation with branching number β=∫ξ(t)dt
    Used to interpret ξ0 as expected offspring per event (Section 2, Eq. 2).
  • domain assumption The StHP intensity in Eq. (20) is a valid conditional intensity
    The background term sums over all N events without restricting t_n < t, so the intensity is not adapted to the past; the paper does not justify this.
  • domain assumption Observation window 2014-2024 with all 412,376 events observed without measurement error or missingness
    The analysis treats Gun Violence Archive records as exact event times and locations (Section 3).
  • domain assumption Spatiotemporal stationarity and spatially uniform background over the whole US
    The model has no population or spatial covariate; any nonstationarity or urban density is absorbed by parameters (Eq. 20).
  • domain assumption MCMC convergence diagnostics indicate samples from the posterior
    Reliance on standard diagnostics (Rhat < 1.003, ESS > 10,000) in Section 3; raw chains are not independently verified.

pith-pipeline@v1.3.0-alltime-deepseek · 13292 in / 13417 out tokens · 115025 ms · 2026-08-01T21:23:04.718976+00:00 · methodology

0 comments
read the original abstract

Hawkes processes (HP) are a large class of stochastic point process models scientists have used to analyze contagion phenomena ranging from earthquakes, infectious diseases and biological neurons to financial trading activity, memes on social media and gun violence. We introduce applications of HP to the latter before reviewing general strategies for fitting HP to data, paying attention to the influence of model structure on computational scalability considerations. We then apply a recently developed high-performance computing powered Bayesian inference strategy for the spatiotemporal HP analysis of 412,376 acts of gun violence in the U.S. between 2014 and 2024. We finish with a discussion of model fit and directions for future research.

Figures

Figures reproduced from arXiv: 2607.16081 by Andrew J. Holbrook, Jasen Zhang, Seyoon Ko.

Figure 1
Figure 1. Figure 1: Temporal (a) and spatial (b) distributions of 412,376 American gun violence events recorded from 2014 to 2024 and provided by the Gun Violence Archive. We use the high￾performance computational inference strategy of Ko et al. (2024) to perform Bayesian infer￾ence and analyze this data in only 10 hours despite the task requiring more than O(16×1010) floating-point operations per iteration of Markov chain Mo… view at source ↗
Figure 2
Figure 2. Figure 2: Absolute and relative speedups for spatiotemporal Hawkes process likelihood [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗

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

63 extracted references · 2 linked inside Pith

  1. [1]

    Bottou, L. (2010). Large-scale machine learning with stochastic gradient descent. In Proceedings of COMPSTAT'2010: 19th International Conference on Computational StatisticsParis France, August 22-27, 2010 Keynote, Invited and Contributed Papers , pp.\ 177--186. Springer

  2. [2]

    Brantingham, P. J., J. Carter, J. MacDonald, C. Melde, and G. Mohler (2021). Is the recent surge in violence in american cities due to contagion? Journal of criminal justice\/ 76 , 101848

  3. [3]

    Br \'e maud, P. and L. Massouli \'e (1996). Stability of nonlinear Hawkes processes . The Annals of Probability\/ 24\/ (3), 1563 -- 1588

  4. [4]

    Chavez-Demoulin, V. and J. McGill (2012). High-frequency financial data modeling using hawkes processes. Journal of Banking & Finance\/ 36\/ (12), 3415--3426

  5. [5]

    Choi, E., N. Du, R. Chen, L. Song, and J. Sun (2015). Constructing disease network and temporal progression model via context-sensitive H awkes process. In 2015 IEEE International Conference on Data Mining , pp.\ 721--726. IEEE

  6. [6]

    Da Fonseca, J. and R. Zaatour (2014). Hawkes process: Fast calibration, application to trade clustering, and diffusive limit. Journal of Futures Markets\/ 34\/ (6), 548--579

  7. [7]

    Daley, D. J. and D. Vere-Jones (2003). Conditional intensities and likelihoods. An Introduction to the Theory of Point Processes: Volume I: Elementary Theory and Methods\/ , 211--287

  8. [8]

    Daw, A. and J. Pender (2023). Matrix calculations for moments of M arkov processes. Advances in Applied Probability\/ 55\/ (1), 126--150

  9. [9]

    Dempster, A. P., N. M. Laird, and D. B. Rubin (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of the royal statistical society: series B (methodological)\/ 39\/ (1), 1--22

  10. [10]

    Cheng, and Y

    Dong, Z., X. Cheng, and Y. Xie (2023). Spatio-temporal point processes with deep non-stationary kernels. In The Eleventh International Conference on Learning Representations

  11. [11]

    Liniger, and L

    Embrechts, P., T. Liniger, and L. Lin (2011). Multivariate hawkes processes: an application to financial data. Journal of Applied Probability\/ 48\/ (A), 367--378

  12. [12]

    Fagan, J., D. L. Wilkinson, and G. Davies (2007). Social Contagion of Violence , pp.\ 688–724. Cambridge Handbooks in Psychology. Cambridge University Press

  13. [13]

    Gelfand, A. and A. Smith (1990). Sampling-based approaches to calculating marginal densities. Journal of the American Statistical Association\/ 85 , 398--409

  14. [14]

    Geman, S. and D. Geman (1984). Stochastic relaxation, gibbs distributions, and the bayesian restoration of images. IEEE Transactions on Pattern Analysis and Machine Intelligence\/ PAMI-6\/ (6), 721--741

  15. [15]

    Hall, A. R. (2005). Generalized method of moments (advanced texts in econometrics series, oxford university press)

  16. [16]

    Hardiman, S. J., N. Bercot, and J.-P. Bouchaud (2013). Critical reflexivity in financial markets: a hawkes process analysis. The European Physical Journal B\/ 86\/ (10), 442

  17. [17]

    Hastings, W. K. (1970). Monte carlo sampling methods using markov chains and their applications. Biometrika\/ 57 , 97--109

  18. [18]

    Hawkes, A. (1972). Spectra of some mutually exciting point processes with associated variables. Stochastic point processes\/ , 261--271

  19. [19]

    Hawkes, A. and L. Adamopoulos (1973). Cluster models for earthquakes-regional comparisons. Bull. Int. Stat. Inst.\/ 45\/ (3), 454--461

  20. [20]

    Hawkes, A. G. (1971a). Point spectra of some mutually exciting point processes. Journal of the Royal Statistical Society: Series B (Methodological)\/ 33\/ (3), 438--443

  21. [21]

    Hawkes, A. G. (1971b). Spectra of some self-exciting and mutually exciting point processes. Biometrika\/ 58\/ (1), 83--90

  22. [22]

    Hawkes, A. G. (2018). Hawkes processes and their applications to finance: a review. Quantitative Finance\/ 18\/ (2), 193--198

  23. [23]

    Hawkes, A. G. and D. Oakes (1974). A cluster process representation of a self-exciting process. Journal of applied probability\/ 11\/ (3), 493--503

  24. [24]

    Hochreiter, S. and J. Schmidhuber (1997). Long short-term memory. Neural Computation\/ 9\/ (8), 1735--1780

  25. [25]

    Holbrook, A. J., X. Ji, and M. A. Suchard (2022a). Bayesian mitigation of spatial coarsening for a hawkes model applied to gunfire, wildfire and viral contagion. The Annals of Applied Statistics\/ 16\/ (1), 573--595

  26. [26]

    Holbrook, A. J., X. Ji, and M. A. Suchard (2022b). From viral evolution to spatial contagion: a biologically modulated hawkes model. Bioinformatics\/ 38\/ (7), 1846--1856

  27. [27]

    Holbrook, A. J., P. Lemey, G. Baele, S. Dellicour, D. Brockmann, A. Rambaut, and M. A. Suchard (2021). Massive parallelization boosts big bayesian multidimensional scaling. Journal of Computational and Graphical Statistics\/ 30\/ (1), 11--24

  28. [28]

    Holbrook, A. J., C. E. Loeffler, S. R. Flaxman, and M. A. Suchard (2021). Scalable bayesian inference for self-excitatory stochastic processes applied to big american gunfire data. Statistics and Computing\/ 31\/ (1), 1--15

  29. [29]

    Kelly, J. D., J. Park, R. J. Harrigan, N. A. Hoff, S. D. Lee, R. Wannier, B. Selo, M. Mossoko, B. Njoloko, E. Okitolonda-Wemakoy, et al. (2019). Real-time predictions of the 2018--2019 ebola virus disease outbreak in the democratic republic of the congo using hawkes point process models. Epidemics\/ 28 , 100354

  30. [30]

    Kim, H. (2011). Spatio-temporal point process models for the spread of avian influenza virus (H5N1) . Ph.\ D. thesis, UC Berkeley

  31. [31]

    Ko, S., M. A. Suchard, and A. J. Holbrook (2024). Scaling Hawkes processes to one million COVID-19 cases . arXiv preprint arXiv:2407.11349\/

  32. [32]

    Laub, P. J., Y. Lee, and T. Taimre (2021). The elements of H awkes processes . Springer

  33. [33]

    Laub, P. J., T. Taimre, and P. K. Pollett (2015). Hawkes processes. arXiv preprint arXiv:1507.02822\/

  34. [34]

    Leskovec, J. and A. Krevl (2014, June). SNAP Datasets : Stanford large network dataset collection. http://snap.stanford.edu/data

  35. [35]

    Levenberg, K. (1944). A method for the solution of certain non-linear problems in least squares. Quarterly of applied mathematics\/ 2\/ (2), 164--168

  36. [36]

    Linderman, S. and R. Adams (2014). Discovering latent network structure in point process data. In International conference on machine learning , pp.\ 1413--1421. PMLR

  37. [37]

    Linderman, S. W., Y. Wang, and D. M. Blei (2017). Bayesian inference for latent hawkes processes. Advances in Neural Information Processing Systems\/

  38. [38]

    Orbanz, Z

    Lloyd, J., P. Orbanz, Z. Ghahramani, and D. M. Roy (2012). Random function priors for exchangeable arrays with applications to graphs and relational data. In F. Pereira, C. Burges, L. Bottou, and K. Weinberger (Eds.), Advances in Neural Information Processing Systems , Volume 25. Curran Associates, Inc

  39. [39]

    Loeffler, C. and S. Flaxman (2018). Is gun violence contagious? a spatiotemporal test. Journal of Quantitative Criminology\/ 34\/ (4), 999--1017

  40. [40]

    Loftin, C. (1986). Assaultive violence as a contagious social process. Bulletin of the New York Academy of Medicine\/ 62\/ (5), 550

  41. [41]

    Marquardt, D. W. (1963). An algorithm for least-squares estimation of nonlinear parameters. Journal of the society for Industrial and Applied Mathematics\/ 11\/ (2), 431--441

  42. [42]

    Mei, H. and J. M. Eisner (2017). The neural hawkes process: A neurally self-modulating multivariate point process. In Advances in Neural Information Processing Systems , pp.\ 6754--6764

  43. [43]

    Metropolis, N., A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller (1953). Equation of state calculations by fast computing machines. The journal of chemical physics\/ 21\/ (6), 1087--1092

  44. [44]

    Held, et al

    Meyer, S., L. Held, et al. (2014). Power-law models for infectious disease spread. The Annals of Applied Statistics\/ 8\/ (3), 1612--1639

  45. [45]

    Mohler, G. (2014). Marked point process hotspot maps for homicide and gun crime prediction in chicago. International Journal of Forecasting\/ 30\/ (3), 491--497

  46. [46]

    Murray, I., R. P. Adams, and D. J. MacKay (2010). Elliptical slice sampling. Journal of Machine Learning Research: Workshop and Conference Proceedings\/ 9 , 541--548

  47. [47]

    Ogata, Y. (1988). Statistical models for earthquake occurrences and residual analysis for point processes. Journal of the American Statistical association\/ 83\/ (401), 9--27

  48. [48]

    Ozaki, T. (1979). Maximum likelihood estimation of hawkes' self-exciting point processes. Annals of the Institute of Statistical Mathematics\/ 31\/ (1), 145--155

  49. [49]

    Park, J., F. P. Schoenberg, A. L. Bertozzi, and P. J. Brantingham (2021). Investigating clustering and violence interruption in gang-related violent crime data using spatial--temporal point processes with covariates. Journal of the American Statistical Association\/ 116\/ (536), 1674--1687

  50. [50]

    Plummer, M., N. Best, K. Cowles, and K. Vines (2006). Coda: Convergence diagnosis and output analysis for mcmc. R News\/ 6\/ (1), 7--11

  51. [51]

    Rasmussen, C. E. and C. K. I. Williams (2005, 11). Gaussian Processes for Machine Learning . The MIT Press

  52. [52]

    Reinhart, A. (2018). A review of self-exciting spatio-temporal point processes and their applications. Statistical Science\/ 33\/ (3), 299--318

  53. [53]

    Mishra, Q

    Rizoiu, M.-A., S. Mishra, Q. Kong, M. Carman, and L. Xie (2018). Sir- H awkes: Linking epidemic models and H awkes processes to model diffusions in finite populations. In Proceedings of the 2018 World Wide Web Conference on World Wide Web , pp.\ 419--428. International World Wide Web Conferences Steering Committee

  54. [54]

    Rubin, I. (1972). Regular point processes and their detection. IEEE Transactions on Information Theory\/ 18\/ (5), 547--557

  55. [55]

    Al Hasan, J

    Sha, H., M. Al Hasan, J. Carter, and G. Mohler (2020). Interpretable hawkes process spatial crime forecasting with tv-regularization. In 2020 IEEE International Conference on Big Data (Big Data) , pp.\ 3228--3236. IEEE

  56. [56]

    RStan : the R interface to Stan

    Stan Development Team (2024). RStan : the R interface to Stan . R package version 2.32.5

  57. [57]

    Taylor, R. M., M. A. Simon, and D. M. Patel (2013). Contagion of violence: Workshop summary . National Academies Press

  58. [58]

    Truccolo, W. (2016). From point process observations to collective neural dynamics: Nonlinear hawkes process glms, low-dimensional dynamics and coarse graining. Journal of Physiology-Paris\/ 110\/ (4), 336--347

  59. [59]

    Veen, A. and F. P. Schoenberg (2008). Estimation of space-time branching process models in seismology using an em--type algorithm. Journal of the American Statistical Association\/ 103\/ (482), 614--624

  60. [60]

    Yang, S.-H. and H. Zha (2013). Mixture of mutually exciting processes for viral diffusion. In International Conference on Machine Learning , pp.\ 1--9

  61. [61]

    Zhao, Q., M. A. Erdogdu, H. Y. He, A. Rajaraman, and J. Leskovec (2015). Seismic: A self-exciting point process model for predicting tweet popularity. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , KDD '15, New York, NY, USA, pp.\ 1513–1522. Association for Computing Machinery

  62. [62]

    Zhu, S. and Y. Xie (2022). Spatiotemporal-textual point processes for crime linkage detection. The Annals of Applied Statistics\/ 16\/ (2), 1151--1170

  63. [63]

    Ogata, and D

    Zhuang, J., Y. Ogata, and D. Vere-Jones (2004). Analyzing earthquake clustering features by using stochastic reconstruction. Journal of Geophysical Research: Solid Earth\/ 109\/ (B5)