Pith. sign in

REVIEW 3 major objections 5 minor 70 references

Anytime-Valid Tests for Sparse Anomalies

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that anytime-valid tests for sparse anomalies have a sharp detection moment: before t*, no test martingale accumulates evidence; after it, the oracle and an adaptive mixture martingale both stop with probability tending…

desk verdict First anytime-valid treatment of sparse anomaly detection with a real oracle threshold theorem; the adaptive construction has a proof typo and a T*<1 gap, but the stress-test's main sign complaint misreads the grid. read the letter →

arxiv 2506.22588 v2 pith:H4XAB7K7 submitted 2025-06-27 math.ST stat.TH

Muriel F. Pérez-Ortiz , Rui M. Castro , Ivo V. Stoepker This is my paper · ORCID
classification math.STstat.TH MSC 62F0362L1060G40
keywords anytime-validtestssparsenormalmeansminimaxhypothesistestingsequentiallikelihoodratiomartingalemixtureSPRThighercriticismdetectionboundary
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper establishes a sharp time threshold for sequentially detecting sparse anomalies among K Gaussian data streams. Under the scaling ε*=$K^{{-β*}}$ and δ*=√(2 ln K / T*), it shows that before t*=T*ρ(β*) no test martingale can accumulate evidence in expected-log terms, while after t* the oracle likelihood-ratio martingale grows without bound and its one-sided SPRT stops with probability tending to one. The main constructive contribution is a mixture likelihood-ratio martingale over a K-dependent grid that attains the same two-sided threshold without knowing ε* or δ*, and can be computed in O($K^{{3/2}}$) operations per time step. Because any anytime-valid test yields a fixed-sample test, this also produces an adaptive fixed-sample test matching the known detection phase transition.

What carries the argument

The central object is the likelihood-ratio test martingale, a nonnegative martingale starting at one under the null; with known parameters it is E_t^*=∏_i[(1−ε*)+ε* p_{δ*}(X_{i,1..t})/p_0(X_{i,1..t})], and the adaptive construction replaces the unknown parameters by a uniform mixture E_t(Π) over a grid G_C. The analysis runs through Kullback-Leibler divergence, second-moment concentration bounds, and the piecewise detection boundary ρ(β), which locates the threshold t*=T*ρ(β*). The grid prior is designed so that, under the misspecification bounds |1−√(T_K/T*)|$K^{{1−β_K}}$=O(1) and |β_K−β*|ln K→0, at least one grid point is close enough to the truth for Proposition 2.4 to transfer the oracle threshold behavior to the mixture martingale.

What would settle it

Take K=$10^{5}$, β*=0.75, T*=40, so ε*=$K^{{-β*}}$ and δ*=√(2 ln K/40), and simulate the adaptive mixture martingale E_t(Π) under P*. The theorems predict P*{τ_Π≤t}→0 for t<t* and →1 for t>t*, with t*=40ρ(0.75)=40(1−√0.25)^2=10; a simulation whose transition occurs at a different time would refute the claimed threshold. A separate check is whether the printed grid contains any δ with 2 ln K/$δ^{2}$>1, as the proof requires.

Watch

Extended reading notes

Core claim

At fixed time t and growing K, the problem is governed by a detection moment t*=T*ρ(β*), where ρ(β*) is the piecewise boundary (β*−1/2 on (1/2,3/4) and (1−√(1−β*))^2 on [3/4,1]). Theorem 2.1 shows that the log-optimal oracle likelihood ratio E_t^* satisfies E*[ln E_t^*]→0 for t<t* and →∞ for t>t*, so no test martingale can have nontrivial expected-log growth before the threshold and the oracle must grow after it. Theorem 2.3 shows the same dichotomy for the stopping time of the oracle one-sided SPRT. Theorem 2.6 and Theorem 2.7 extend both statements to the adaptive mixture martingale E_t(Π), a uniform mixture of likelihood ratios over a grid of (ε,δ) values, which achieves the same threshold behavior without knowing the true parameters. The paper claims the sequential version of sparse-anomaly testing has its own phase transition, related but not implied by the fixed-sample transition.

Load-bearing premise

The adaptive guarantee rests on the K-dependent grid prior Π=Uniform(G_C) containing a point whose misspecification bias satisfies |1−√(T_K/T*)|$K^{{1−β_K}}$=O(1) and |β_K−β*| ln K→0, with T*<C. The paper's displayed grid G_{δ,C}(ε) uses δ_i=√(2 ln K exp(i/⌈εK⌉)), which places T_i=2 ln K/$δ_i^{2}$ in [1/C,1], while the proof of Theorem 2.6 requires grid values with T_i up to C; read literally, the displayed grid cannot cover T*>1 and the stated regime of Theorem 2.6 is not established.

Editorial extensions

If this is right

  • Continuous monitoring of many streams for sparse anomalies acquires a sharp detection moment: before t*, no test martingale can have nontrivial expected-log growth, so early stopping is genuinely impossible, not merely hard.
  • After t*, the mixture SPRT stops with probability tending to one while retaining type-I error control at every stopping time, so practitioners can monitor indefinitely without a time-horizon-dependent Bonferroni penalty.
  • The mixture likelihood ratio yields a fixed-sample adaptive test whose power transitions at the same boundary as the oracle likelihood ratio and the higher-criticism test, as a direct by-product.
  • The O(K^{3/2} ln K) per-time-step computational cost makes the adaptive test feasible for large K, and the cost depends only logarithmically on the horizon bound C.
  • Finite-K simulations show the low-to-high power transition sharpening as K grows, consistent with the asymptotic dichotomy.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The proof's grid as displayed appears to cover only T_i≤1, while Theorem 2.6 needs grid points with T_i up to C; if the displayed formula is literal, the adaptive guarantee for T*>1 is an unproven step rather than a closed result.
  • The threshold structure suggests a practical design rule: choose a desired sparsity-sensitivity pair (β*,T*), read off t*=T*ρ(β*) as the promised detection time, and use the mixture martingale without estimating parameters.
  • The detection-but-not-identification window between t* and T* implies that early rejection can be trusted as evidence that anomalies exist, but not as a list of which streams are anomalous; identification would require the stronger regime r>β.
  • A natural extension is to test whether the same threshold transfers to rank-based or distribution-free stream statistics; the Gaussian machinery suggests controlling a Kullback-Leibler-type divergence is the key ingredient, but that transfer is not shown in the paper.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper develops anytime-valid (AV) tests for detecting sparse anomalies among K Gaussian streams, under the contamination model where each stream is anomalous with probability ε and has mean shift δ. The authors analyze the oracle likelihood-ratio martingale E*_t and show that, for ε*=K^{-β*} and δ*=sqrt(2(1/T*) ln K), the expected log-growth E*[ln E*_t] transitions from 0 to infinity at t*=T*ρ(β*), and that the corresponding SPRT stops with probability tending to 0 before t* and to 1 after t* (Theorems 2.1 and 2.3). They then construct an adaptive mixture martingale E_t(Π) based on a K-dependent grid prior and claim that it attains the same threshold behavior without knowing (ε*,δ*) (Theorems 2.6 and 2.7). Numerical simulations compare the proposed tests with fixed-sample likelihood-ratio and higher-criticism benchmarks.

Significance. If fully established, the paper would be a substantial contribution: it transfers the fixed-sample sparse-detection phase transition into a sequential anytime-valid setting, provides an explicit and computationally tractable adaptive construction (O(K^{3/2} ln K) per step), and backs the theory with finite-K simulations. The oracle analysis is largely self-contained, and the lower-bound part legitimately imports the fixed-sample HC risk result rather than assuming the desired conclusion. The adaptive construction is a genuine methodological advance, and the Monte Carlo calibration of thresholds is a reasonable and clearly described practical choice. The main caveat is that the adaptive threshold theorems are not established as printed because the displayed grid in Eq. (9) is inconsistent with the proof of Theorem 2.6 and does not cover the full stated parameter range.

major comments (3)
  1. [Eq. (9) and proof of Theorem 2.6 (Section B.4)] As printed, the grid in Eq. (9) reads δ_i = sqrt(2 ln K exp(i/⌈εK⌉)) (or equivalently sqrt(2 ln K) exp(i/(2⌈εK⌉))), so that T_i = 2 ln K / δ_i^2 = exp(-i/⌈εK⌉) lies in [1/C,1]. The proof of Theorem 2.6, however, uses T_i = exp(i/⌈Kε°⌉) and needs a grid point with T° ≤ T* and T* < C. For T* > 1 the displayed grid cannot provide such a point, condition (7) of Proposition 2.4 cannot be verified, and Theorem 2.6 is not established for the stated regime. This is a load-bearing definition-proof mismatch, though it appears correctable by writing the intended grid as δ_i = sqrt(2 ln K / exp(i/⌈εK⌉)).
  2. [Proof of Theorem 2.6 (Section B.4)] The bracketing inequality in the proof is also reversed. The proof states δ_i < δ* ≤ δ_{i+1} with T_i = exp(i/⌈Kε°⌉) and T_{i+1} = exp((i+1)/⌈Kε°⌉). Since T_i is increasing in i, δ_i = sqrt(2 ln K / T_i) is decreasing in i, so the displayed ordering is impossible for the claimed choice T°=T_i ≤ T*. The intended ordering should be δ_{i+1} < δ* ≤ δ_i, which is what would make T_i ≤ T* and d_T = 1 - sqrt(T_i/T*) nonnegative and of order K^{-(1-β°)}.
  3. [Theorem 2.6 and Theorem 2.7, parameter range T*<C] Even after correcting the sign in Eq. (9), the proposed grid with T_i = exp(i/⌈εK⌉) and i ≥ 1 covers only T° ≥ exp(1/⌈εK⌉) ≈ 1. The theorems assume only T* < C with C > 1 and allow 0 < T* < 1. For such T* no grid point can satisfy T° ≤ T* together with |d_T| K^{1-β°} = O(1), so the proof of Theorem 2.6 (and hence Theorem 2.7, which relies on the same grid point) does not cover the full stated range. The statements should either restrict T* to T* > 1 or extend the grid to cover T° ∈ [1/C,C].
minor comments (5)
  1. [Section 1.3, paragraph after Theorem 2.1] The text defines β* = ln(1/ε*), but since ε* = K^{-β*}, this should read β* = ln(1/ε*)/ln K.
  2. [Definition 2.5] The definition says 'Let C > 0', but the grid construction and Theorem 2.6 require C > 1; also the K-dependence of G_C and Π is suppressed, which can confuse the reader. Writing G_C(K) and Π_K would clarify the construction.
  3. [Proof of Theorem 2.7, Eq. (56)] Eq. (56) writes P*{τ* ≤ t} = P*{max_{s≤t} E_s(Π) ≥ 1/α}, but τ* was defined earlier for the oracle SPRT; the stopping time here should be τ_Π.
  4. [Theorem 2.7] The theorem states α ∈ (0,1], while all other level statements use α ∈ (0,1); this should be harmonized.
  5. [Section 2.2, opening paragraph] There is a typo 'As we will wee' that should read 'As we will see'.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the sequential threshold is benchmarked against an external fixed-sample boundary and proven by direct KL computations; the only flagged issue is an Eq. (9) grid sign mismatch, which is a correctness gap in the proof of Theorem 2.6, not a circular reduction.

full rationale

The derivation chain is not circular. Theorem 2.1 benchmarks the oracle likelihood-ratio martingale against the externally established fixed-sample detection boundary rho(beta) from [27,15], and the paper explicitly states that the sequential threshold is related to but not implied by the fixed-sample phase transition. The upper-bound direction (Lemma B.1) is a direct truncation computation of E-star[ln E_t*], while the lower bound (Lemma B.2) bridges an external power result to the KL divergence via the inequality (18), KL >= -ln(4R(psi)), a genuine implication, not a restatement. Corollary 2.2 then transfers the oracle result to all test martingales through the log-optimality theorem of [33], an external result. The adaptive construction (Definition 2.5) is a fixed K-dependent uniform grid prior; the misspecification window (7)-(8) is derived with explicit rates in Proposition B.3 from direct Gaussian integral computations (Lemmas B.5-B.9, B.12), and the grid is engineered so that for any true (beta*,T*) with T*<C some grid point lands in the window. No parameter is fitted to data to obtain the theorems; the Monte Carlo thresholds A* and A_Pi (Sections 3 and D) calibrate only the level in simulations and are not part of the theoretical claims. Self-citations [46,47] appear only as background on sparse-mixture detection and never as load-bearing inputs to the proofs; no uniqueness theorem is imported from the authors' own work. The one flagged issue is a correctness gap, not circularity: as printed, Eq. (9) gives delta_i = sqrt(2 lnK exp(i/ceil(epsilon K))), hence T_i = exp(-i/ceil(epsilon K)) in [1/C,1], whereas the proof of Theorem 2.6 in Section B.4 requires T_i = exp(i/ceil(K epsilon^o)) to bracket T* up to C; taken literally, no grid point can satisfy condition (7) for T*>1, so Theorem 2.6 is not established for the stated regime. This is a definition-proof sign mismatch (apparently typographical), not a circular reduction, and it does not raise the circularity score.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central results rest on the Gaussian contamination model, the external fixed-sample detection boundary rho used as the benchmark, the martingale toolkit for anytime validity, and a customized grid-closeness condition for the adaptive construction. The grid-closeness condition is the only assumption ad hoc to this paper; the rest are standard parts of the relevant literatures.

free parameters (2)
  • C (grid and horizon parameter) = C = 5 T* = 200 in simulations; user-specified in the theory
    Defines the mixture prior support: Theorem 2.6 assumes T* < C, and the grid size and computational cost scale logarithmically with C. It is not fitted to data, but it is a tuning choice.
  • Monte Carlo calibrated thresholds A* and A_Pi = estimated from 10^4 null samples; individual values not reported
    Used in simulations to achieve an exact 5% level instead of the conservative 1/alpha threshold. These are fitted to null simulations and do not enter the main theoretical claims.
assumptions (6)
  • domain assumption Contamination model (1): A_i ~ Bernoulli(eps), X_{i,t} | A_i ~ Normal(delta A_i, 1), independent across streams and time.
    All tests are designed and analyzed under this model; any power claim is relative to it.
  • standard math Fixed-sample sparse detection phase transition (Ingster, Donoho-Jin): tests are powerless when r < rho(beta) and Higher Criticism tests have risk tending to zero when r > rho(beta).
    Used in Lemma B.2 and Theorem 2.3 to transfer the boundary rho to the KL divergence and stopping time; not proved in this paper.
  • standard math Log-optimality of the likelihood ratio test martingale under a simple alternative (Kelly, Koolen-Grünwald).
    Justifies E_t^* in Eq. (4) as the oracle benchmark and supports Definition 1.3.
  • standard math Ville's inequality and the construction of anytime-valid tests from test (super)martingales.
    Basis of Definitions 1.1 and 1.2; used for type-I error control.
  • ad hoc to paper Grid closeness condition (7)-(8): the prior grid must contain parameters satisfying |d_T| K^{1-beta}=O(1) and |d_beta| lnK -> 0.
    This is a load-bearing assumption specific to the adaptive construction; Proposition 2.4 requires it and the grid is designed to satisfy it.
  • standard math Renyi divergence lower bound d_TV^2 (van Erven and Harremoes, Theorem 31) and Cramer-Chernoff concentration bounds.
    Used in the proof of Theorem 2.3 to show that the oracle SPRT stops with probability tending to one above t*.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Anytime-Valid Tests for Sparse Anomalies." pith.science (2026). https://pith.science/paper/H4XAB7K7

@misc{pith2026250622588,
  author       = {Pith},
  title        = {Pith review of: Anytime-Valid Tests for Sparse Anomalies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H4XAB7K7}},
  note         = {Machine review of arXiv:2506.22588}
}
read the original abstract

We consider the problem of testing sequentially for the presence of sparse anomalies among a large number of data streams. To this end, we design and analyze Anytime-Valid (AV) tests, which retain type-I error control at arbitrary stopping times. Existing results address exclusively the nonsequential case, which exhibits a subtle phase transition between two regimes where tests are either powerless or powerful. In our sequential setting, we argue, two challenges arise: (1) the standard analysis of AV tests cannot be executed in the relevant sample-size regime; and (2) standard constructions of parameter-adaptive AV tests are either analytically intractable or computationally unfeasible. This work addresses these challenges. Borrowing insights from the nonsequential literature, we propose a framework to analyze AV tests and their shortest possible sample sizes. Under this framework, we show that, in the Gaussian location setting, the oracle AV test has a delicate threshold behavior that is related to -- but not implied by -- the phase transition observed in optimal nonsequential tests. Our main results include a computationally efficient, parameter-adaptive AV test; we show that it achieves the same threshold behavior as the oracle AV test. Numerical simulations illustrate these theoretical findings.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 51 canonical work pages

  1. [1]

    and STEGUN, I

    ABRAMOWITZ, M. and STEGUN, I. A. (1965).Handbook of Mathematical Functions: With Formulas, Graphs, and Mathematical Tables. Courier Corporation Google-Books-ID: MtU8uP7XMvoC. ANYTIME-V ALID TESTS FOR SPARSE ANOMALIES21

    work page 1965

  2. [2]

    and LUGOSI, G

    ADDARIO-BERRY, L., BROUTIN, N., DEVROYE, L. and LUGOSI, G. (2010). On Combinatorial Testing Problems.The Annals of Statistics383063–3092. Publisher: Institute of Mathematical Statistics

    work page 2010

  3. [3]

    and CHEN, S

    ARIAS-CASTRO, E. and CHEN, S. (2017). Distribution-free multiple testing.Electronic Journal of Statis- tics111983 – 2001. https://doi.org/10.1214/17-EJS1277

    work page doi:10.1214/17-ejs1277 2017

  4. [4]

    and YING, A

    ARIAS-CASTRO, E. and YING, A. (2019). Detection of sparse mixtures: higher criticism and scan statis- tic.Electronic Journal of Statistics13208–230. Publisher: Institute of Mathematical Statistics and Bernoulli Society. https://doi.org/10.1214/18-EJS1512

    work page doi:10.1214/18-ejs1512 2019

  5. [5]

    and PANARETOS, J

    BERSIMIS, S., PSARAKIS, S. and PANARETOS, J. (2007). Multivariate statistical process control charts: an overview.Quality and Reliability Engineering International23517-543. https://doi.org/10.1002/qre. 829

    work page doi:10.1002/qre 2007

  6. [6]

    and WANG, R

    BLIER-WONG, C. and WANG, R. (2025). Improved thresholds for e-values. arXiv:2408.11307 [math]. https://doi.org/10.48550/arXiv.2408.11307

    work page doi:10.48550/arxiv.2408.11307 2025

  7. [7]

    BREIMAN, L. (1961). Optimal Gambling Systems for Favorable Games.Proceedings of the Fourth Berke- ley Symposium on Mathematical Statistics and Probability, Volume 1: Contributions to the Theory of Statistics4.165–79. Publisher: University of California Press

    work page 1961

  8. [8]

    T., JESSIEJENG, X

    CAI, T. T., JESSIEJENG, X. and JIN, J. (2011). Optimal Detection of Heterogeneous and Heteroscedastic Mixtures.Journal of the Royal Statistical Society Series B: Statistical Methodology73629–662. https: //doi.org/10.1111/j.1467-9868.2011.00778.x

    arXiv 2011

Show all 70 references
  1. [9]

    CAI, T. T. and JIN, J. (2010). Optimal rates of convergence for estimating the null density and proportion of nonnull effects in large-scale multiple testing.The Annals of Statistics38100–145. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/09-AOS696

    2010 doi

  2. [10]

    T., JIN, J

    CAI, T. T., JIN, J. and LOW, M. G. (2007). Estimation and confidence sets for sparse normal mixtures.The Annals of Statistics352421–2449. Publisher: Institute of Mathematical Statistics. https://doi.org/10. 1214/009053607000000334

    2007

  3. [11]

    CAI, T. T. and WU, Y. (2014). Optimal Detection of Sparse Mixtures Against a Given Null Distribution. IEEE Transactions on Information Theory602217–2232. Conference Name: IEEE Transactions on Information Theory. https://doi.org/10.1109/TIT.2014.2304295

    2014

  4. [12]

    and VAART, A.V.D

    CASTILLO, I. and VAART, A.V.D. (2012). Needles and Straw in a Haystack: Posterior concentration for possibly sparse sequences.The Annals of Statistics402069–2101. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/12-AOS1029

    2012 doi

  5. [13]

    H., LAI, T

    DELAPEÑA, V. H., LAI, T. L. and SHAO, Q.-M. (2009).Self-Normalized Processes.Probability and its Applications. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-85636-8

    2009 doi

  6. [14]

    P., LAIRD, N

    DEMPSTER, A. P., LAIRD, N. M. and RUBIN, D. B. (1977). Maximum Likelihood from Incomplete Data Via the EM Algorithm.Journal of the Royal Statistical Society: Series B (Methodological)391–22. https://doi.org/10.1111/j.2517-6161.1977.tb01600.x

    1977

  7. [15]

    and JIN, J

    DONOHO, D. and JIN, J. (2004). Higher criticism for detecting sparse heterogeneous mixtures.The An- nals of Statistics32962–994. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/ 009053604000000265

    2004

  8. [16]

    and JIN, J

    DONOHO, D. and JIN, J. (2015). Higher Criticism for Large-Scale Inference, Especially for Rare and Weak Effects.Statistical Science301–25. Publisher: Institute of Mathematical Statistics. https://doi.org/10. 1214/14-STS506

    2015

  9. [17]

    and SZABÓ, B

    ERVEN, T.V. and SZABÓ, B. (2021). Fast Exact Bayesian Inference for Sparse Signals in the Normal Se- quence Model.Bayesian Analysis16933–960. Publisher: International Society for Bayesian Analysis. https://doi.org/10.1214/20-BA1227

    2021 doi

  10. [18]

    and RAMDAS, A

    FISCHER, L. and RAMDAS, A. (2025). Improving Wald’s (approximate) sequential probability ratio test by avoiding overshoot. arXiv:2410.16076 [stat]. https://doi.org/10.48550/arXiv.2410.16076

    2025 doi arXiv

  11. [19]

    and KOOLEN, W

    GRÜNWALD, P.,DEHEIDE, R. and KOOLEN, W. (2024). Safe Testing.Journal of the Royal Statistical Society Series B: Statistical Methodologyqkae011. https://doi.org/10.1093/jrsssb/qkae011

    2024 doi

  12. [20]

    GRÜNWALD, P. D. (2007).The Minimum Description Length Principle.Adaptive Computation and Ma- chine Learning series

    2007

  13. [21]

    GRÜNWALD, P. D. (2024). Beyond Neyman–Pearson: E-values enable hypothesis testing with a data-driven alpha.Proceedings of the National Academy of Sciences121e2302098121. https://doi.org/10.1073/ pnas.2302098121

    2024

  14. [22]

    GRÜNWALD, P. D. and ROOS, T. (2019). Minimum description length revisited.International Journal of Mathematics for Industry111930001. Publisher: World Scientific Publishing Co. https://doi.org/10. 1142/S2661335219300018

    2019

  15. [23]

    and FRIEDMAN, J

    HASTIE, T., TIBSHIRANI, R. and FRIEDMAN, J. (2009).The Elements of Statistical Learning: Data Min- ing, Inference, and Prediction, Second Edition, 2 ed.Springer Series in Statistics. Springer-Verlag, New York. 22M.F. PÉREZ-ORTIZ ET AL

    2009

  16. [24]

    HAUPT, J., CASTRO, R. M. and NOWAK, R. (2011). Distilled sensing: Adaptive sampling for sparse detec- tion and estimation.IEEE Transactions on Information Theory576222–6235

    2011

  17. [25]

    R., RAMDAS, A., MCAULIFFE, J

    HOWARD, S. R., RAMDAS, A., MCAULIFFE, J. and SEKHON, J. (2018). Uniform, nonparametric, non- asymptotic confidence sequences.arXiv:1810.08240 [math, stat]. arXiv: 1810.08240

    2018 arXiv

  18. [26]

    and RAMDAS, A

    IGNATIADIS, N., WANG, R. and RAMDAS, A. (2024). E-values as unnormalized weights in multiple test- ing.Biometrika111417–439. https://doi.org/10.1093/biomet/asad057

    2024 doi

  19. [27]

    INGSTER, Y. I. (1994). Minimax detection of a signal in lp metrics.Journal of Mathematical Sciences68 503–515. https://doi.org/10.1007/BF01254275

    1994 doi

  20. [28]

    INGSTER, Y. I. (1997). Some problems of hypothesis testing leading to infinitely divisible distributions. Mathematical Methods of Statistics647–69

    1997

  21. [29]

    INGSTER, Y. I. (2002). Adaptive Detection of a Signal of Growing Dimension. II.Mathematical Methods of Statistics1137–68. OCLC: 196146366

    2002

  22. [30]

    and WALSH, D

    JOHARI, R., KOOMEN, P., PEKELIS, L. and WALSH, D. (2022). Always Valid Inference: Continuous Monitoring of A/B Tests.Operations Research701806–1821. Publisher: INFORMS. https://doi.org/ 10.1287/opre.2021.2135

    2022

  23. [31]

    JOHNSTONE, I. M. and SILVERMAN, B. W. (2004). Needles and straw in haystacks: Empirical Bayes estimates of possibly sparse sequences.The Annals of Statistics321594–1649. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/009053604000000030

    2004 doi

  24. [32]

    KELLYJR., J. L. (1956). A New Interpretation of Information Rate.Bell System Technical Journal35917–

    1956

  25. [33]

    KOOLEN, W. M. and GRÜNWALD, P. (2022). Log-optimal anytime-valid E-values.International Journal of Approximate Reasoning14169–82. https://doi.org/10.1016/j.ijar.2021.09.010

    2022 doi

  26. [34]

    and MOSTASHARI, F

    KULLDORFF, M., HEFFERNAN, R., HARTMAN, J., ASSUNCAO, R. and MOSTASHARI, F. (2005). A Space-Time Permutation Scan Statistic for Disease Outbreak Detection.PLOS Medicine2216

    2005

  27. [35]

    LAI, T. L. (1977). Power-One Tests Based on Sample Sums.The Annals of Statistics5866–880. Publisher: Institute of Mathematical Statistics

    1977

  28. [36]

    and ZHANG, X

    LI, G. and ZHANG, X. (2025). A note on e-values and multiple testing.Biometrika112asae050. https: //doi.org/10.1093/biomet/asae050

    2025 doi

  29. [37]

    and SIEGMUND, D

    LI, J. and SIEGMUND, D. (2015). Higher criticism:p-values and criticism.The Annals of Statistics43 1323–1350. https://doi.org/10.1214/15-AOS1312

    2015 doi

  30. [38]

    and RICE, J

    MEINSHAUSEN, N. and RICE, J. (2006). Estimating the proportion of false null hypotheses among a large number of independently tested hypotheses.The Annals of Statistics34373–393. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/009053605000000741

    2006 doi

  31. [39]

    and STEWART, M

    PORTER, T. and STEWART, M. (2020). Beyond HC: More sensitive tests for rare/weak alternatives.The Annals of Statistics482230–2252. Publisher: Institute of Mathematical Statistics. https://doi.org/10. 1214/19-AOS1885

    2020

  32. [40]

    and SHAFER, G

    RAMDAS, A., GRÜNWALD, P., VOVK, V. and SHAFER, G. (2023). Game-Theoretic Statistics and Safe Anytime-Valid Inference.Statistical Science38576–601. Publisher: Institute of Mathematical Statis- tics. https://doi.org/10.1214/23-STS894

    2023 doi

  33. [41]

    and KOOLEN, W

    RAMDAS, A., RUF, J., LARSSON, M. and KOOLEN, W. (2020). Admissible anytime-valid sequential in- ference must rely on nonnegative martingales.arXiv:2009.03167 [math, stat]. arXiv: 2009.03167

    2020 arXiv

  34. [42]

    and WANG, R

    RAMDAS, A. and WANG, R. (2025). Hypothesis Testing with E-values.Foundations and Trends in Statistics 11–390. https://doi.org/10.1561/3600000002

    2025 doi

  35. [43]

    ROBBINS, H. (1970). Statistical Methods Related to the Law of the Iterated Logarithm.The Annals of Mathematical Statistics411397–1409. Publisher: Institute of Mathematical Statistics. https://doi.org/ 10.1214/aoms/1177696786

    1970

  36. [44]

    and SIEGMUND, D

    ROBBINS, H. and SIEGMUND, D. (1974). The Expected Sample Size of Some Tests of Power One.The Annals of Statistics2415–436. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/ aos/1176342704

    1974

  37. [45]

    (1985).Sequential Analysis.Springer Series in Statistics

    SIEGMUND, D. (1985).Sequential Analysis.Springer Series in Statistics. Springer, New York, NY . https: //doi.org/10.1007/978-1-4757-1862-1

    1985 doi

  38. [46]

    V., CASTRO, R

    STOEPKER, I. V., CASTRO, R. M. and ARIAS-CASTRO, E. (2025). Sparse anomaly detection across referentials: A rank-based higher criticism approach.The Annals of Statistics53676 – 702. https: //doi.org/10.1214/24-AOS2477

    2025 doi

  39. [47]

    V., CASTRO, R

    STOEPKER, I. V., CASTRO, R. M., ARIAS-CASTRO, E. andVAN DENHEUVEL, E. (2024). Anomaly De- tection for a Large Number of Streams: A Permutation-Based Higher Criticism Approach.Journal of the American Statistical Association119461–474. https://doi.org/10.1080/01621459.2022.2126361

    2024

  40. [48]

    and BASSEVILLE, M

    TARTAKOVSKY, A., NIKIFOROV, I. and BASSEVILLE, M. (2014).Sequential Analysis: Hypothesis Testing and Changepoint Detection. Chapman and Hall/CRC, New York. https://doi.org/10.1201/b17279 ANYTIME-V ALID TESTS FOR SPARSE ANOMALIES23

    2014 doi

  41. [49]

    TSYBAKOV, A. B. (2009).Introduction to Nonparametric Estimation.Springer Series in Statistics. Springer, New York, NY . https://doi.org/10.1007/b13794 [50]VANERVEN, T. and HARREMOËS, P. (2014). Rényi Divergence and Kullback-Leibler Divergence.IEEE Transactions on Information T...

    2009

  42. [51]

    (1939).Étude Critique de la Notion de Collectif.Thèses de l’entre-deux-guerres218

    VILLE, J. (1939).Étude Critique de la Notion de Collectif.Thèses de l’entre-deux-guerres218

    1939

  43. [52]

    and WANG, R

    VOVK, V. and WANG, R. (2021). E-values: Calibration, combination and applications.The Annals of Statistics491736–1754. Publisher: Institute of Mathematical Statistics. https://doi.org/10.1214/ 20-AOS2020

    2021

  44. [53]

    and WANG, R

    VOVK, V. and WANG, R. (2024). Nonparametric E-tests of Symmetry.The New England Journal of Statis- tics in Data Science2261–270. Publisher: New England Statistical Society. https://doi.org/10.51387/ 24-NEJSDS60

    2024

  45. [54]

    (1947).Sequential Analysis

    WALD, A. (1947).Sequential Analysis. J. Wiley & Sons, New York

    1947

  46. [55]

    and WOLFOWITZ, J

    WALD, A. and WOLFOWITZ, J. (1948). Optimum Character of the Sequential Probability Ratio Test.The Annals of Mathematical Statistics19326–339. Publisher: Institute of Mathematical Statistics. https: //doi.org/10.1214/aoms/1177730197

    1948

  47. [56]

    WANG, H., and RAMDAS, A. (2025). Anytime-valid t-tests and confidence sequences for Gaussian means with unknown variance.Sequential Analysis4456–110. https://doi.org/10.1080/07474946. 2024.2428245

    2025

  48. [57]

    and RAMDAS, A

    WAUDBY-SMITH, I. and RAMDAS, A. (2023). Estimating means of bounded random variables by betting. Journal of the Royal Statistical Society Series B: Statistical Methodologyqkad009. https://doi.org/10. 1093/jrsssb/qkad009

    2023

  49. [58]

    and RAMDAS, A

    XU, Z. and RAMDAS, A. (2024). Online multiple testing with e-values. InProceedings of The 27th Inter- national Conference on Artificial Intelligence and Statistics3997–4005. PMLR ISSN: 2640-3498

    2024

  50. [59]

    and MIHELIC, J

    ZAGAR, J. and MIHELIC, J. (2022). Big data collection in pharmaceutical manufacturing and its use for product quality predictions.Scientific Data 2022 9:191–11. https://doi.org/10.1038/ s41597-022-01203-x

    2022

  51. [60]

    detection boundary

    ZAMBA, K. D. and TSIAMYRTZIS, P. (2021). Sequential detection framework for real-time biosurveillance based on Shiryaev-Roberts procedure with illustrations using COVID-19 incidence data.Sequential Analysis40149–169. https://doi.org/10.1080/07474946.2021.1912503 APPENDIX A: LE...

    2021

  52. [61]

    AsK→ ∞, Φ(x)∼ 1√ 2π K−ξ √2ξlnK

  53. [62]

    Ifζ K =o((lnK) −1/2)asK→ ∞, then Φ(x+ζ K)∼ Φ(x)in the same limit

  54. [63]

    Then, asK→ ∞, bothΦ(∆, x+ ∆)∼ Φ(∆)and Φ(−∆, x−∆)∼    1√ 2π K −(r1/2 −ξ1/2 )2 √ 2(r1/2−ξ1/2)2 lnK ifξ < r; 1/2ifξ=r; 1ifξ > r

    Let∆ = √ 2rlnK. Then, asK→ ∞, bothΦ(∆, x+ ∆)∼ Φ(∆)and Φ(−∆, x−∆)∼    1√ 2π K −(r1/2 −ξ1/2 )2 √ 2(r1/2−ξ1/2)2 lnK ifξ < r; 1/2ifξ=r; 1ifξ > r

  55. [64]

    For eacha∈R, we have that R ∞ 0 wφ(w−a)dw=φ(a) +aΦ(a)and consequently, if a→ ∞, Z ∞ 0 wφ(w−a)dw∼aand Z ∞ 0 wφ(w+a)dw=o(1)

  56. [65]

    Consequently, ifa→ ∞ Z ∞ 0 w2φ(w−a)dw∼a 2 and Z ∞ 0 w2φ(w+a)dw∼ Φ(a)

    For eacha∈R, R ∞ 0 w2φ(w−a)dw= (a 2 + 1)Φ(a) + a√ 2π e−a2/2. Consequently, ifa→ ∞ Z ∞ 0 w2φ(w−a)dw∼a 2 and Z ∞ 0 w2φ(w+a)dw∼ Φ(a). ANYTIME-V ALID TESTS FOR SPARSE ANOMALIES43 PROOF OFLEMMAB.10. Item 1 is known [1, Formula 7.1.13]. In our notation, For- mula 7.1.13 [1] is equiv...

  57. [66]

    With this lemma at hand and the discussion before it, we can complete the proof of Theorem 2.7 for the case thatt > t ⋆

    the sequenceK7→b K grows faster thanK7→ln(1/(απ K))in the sense that, asK→ ∞, bK ln(1/(απK )) → ∞; and 2.P ⋆{lnE t(ε, δ)≥bK} →1, also asK→ ∞. With this lemma at hand and the discussion before it, we can complete the proof of Theorem 2.7 for the case thatt > t ⋆. Indeed, letb K...

  58. [67]

    fori∈[K], computeπ i,l = ˆεl−1 exp(ˆδl−1zi−ˆδ2 l−1/2) 1−ˆεl−1+ˆεl−1 exp(ˆδl−1zi−ˆδ2 l−1/2)

  59. [68]

    compute ˆδl = P i∈[K] πi,l zi

  60. [69]

    computeˆεl = 1 K P i∈[K] πi,l

  61. [70]

    if|ln ˆεl −ln ˆεl−1|<10 −4 and| ˆδl − ˆδl−1|<10 −4, stop and return(ˆεl, ˆδl). ANYTIME-V ALID TESTS FOR SPARSE ANOMALIES49 If the algorithm fails to converge afterm max steps, we use(ˆε, ˆδ) = (0.0,1.0)—in this case the multiplicative increment ofM t takes the value1. In our e...

  62. [926]

    https:// doi.org/10.1002/j.1538-7305.1956.tb03809.x

    _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/j.1538-7305.1956.tb03809.x. https:// doi.org/10.1002/j.1538-7305.1956.tb03809.x

    1956

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.