betaFeedback

WorksMathematics

Exhaustive adjustment-set selection and causal effects for distribution-valued outcomes, with inference under latent Gaussian spatio-temporal dependence

Abstract

Selecting adjustment covariates for a distribution-valued outcome must preserve dependence that scalar summaries can discard, and surveillance outcomes are also dependent across space and time. Under unconfoundedness and a relative-overlap condition we identify average treatment effects, stochastic and incremental intervention effects, and natural direct and indirect effects for such outcomes, with doubly and triply robust scores. For independent data, adjustment sets are selected exhaustively by distance-covariance sufficient dimension reduction: we give a root-n projector expansion, a basis-invariant criterion tensor, and a simultaneous outer confidence collection calibrated by fractional sampling with a finite-population correction. For dependence, units are functions of a latent Gaussian field, exponential in space and Ornstein-Uhlenbeck in time, plus independent noise. Such data are ρρ-mixing at an exponential rate, and the covariance between any two square-integrable scores is a power series in the latent correlation with positive semidefinite coefficients. Consequences include a central limit theorem for all the causal estimators, the exact undercoverage of independence-based intervals, a consistent Mehler covariance estimator under finite chaos order, and an asymptotically conservative convex-chord estimator; block fractional sampling extends the collection under an assumed linearization. In simulations, fractional sampling raises weak-signal collection coverage from .69-.78 to at least .98 while retaining nearly all candidate sets, and the dependence-aware estimators substantially reduce undercoverage, although none is exact when few units fall within a correlation range. A descriptive analysis of U.S. syphilis surveillance and Medicaid expansion illustrates the estimands and compares dependence-aware with conventional inference.

Introduction

A distribution-valued response can depend on covariates through location, scale or shape, so an adjustment set adequate for its mean need not remove confounding for the distribution itself. For example, the average time of incidence can be unchanged while its temporal concentration differs between exposure groups. Public-health surveillance adds a second difficulty: jurisdictions are neighbours, policies cluster geographically and incidence is persistent, so the independent-sampling approximations behind standard errors and resampling calibrations can fail. Luo et al. [1] introduced exhaustive identification of sufficient adjustment sets through outcome and treatment central subspaces; we build on that criterion and do not propose exhaustive selection as new. Distance-covariance sufficient dimension reduction (SDR) [2, 3, 4, 5] supplies natural estimators when scalar inverse-regression summaries are inadequate. Its root-nn theory for ordinary responses is due to Sheng and Yin [2, 3]. Our extension concerns powered and feature-distance responses, a whitening-aware projector expansion, and simultaneous inference on the adjustment collection. The contributions are as follows.

  • (i) A basis-invariant criterion tensor and its calibration. The criterion is written through products of projectors and residual covariance matrices, a smooth vector-valued object even when its norm vanishes. Conventional subsampling with b=n0.7b = n^{0.7} undercovers badly in a weak-signal design, because of the missing finite-population factor and very small subsamples. We prove that fractional sampling (b/n→c∈(0,1)b/n \to c \in(0, 1)) with the factor {bn/(n−b)}1/2\{bn/(n - b)\}^{1/2} is valid for independent data (Theorem 5). In the same design it restores coverage, partly through the finite-population factor and partly through larger subsets, but the resulting collection is conservative: it protects against excluding a sufficient set rather than discriminating among sets.

  • (ii) A latent Gaussian dependence model. Units are functions of a latent Gaussian field with exponential (Matérn-12\frac{1}{2}) spatial correlation and, for trajectories, Ornstein–Uhlenbeck (OU) temporal correlation, plus independent noise. Such data are ρ\rho-mixing at an exponential rate uniformly in set size (Lemma 1), and the covariance of any two square-integrable scores is a power series ∑mcmMm\sum_m c^m M_m in the latent correlation cc with positive semidefinite coefficients (Proposition 4). This gives the exact undercoverage of independence-based intervals.

  • (iii) Dependence-aware inference. We prove a central limit theorem for augmented inverse-probability-weighted (AIPW) functional contrasts with parametric nuisances (Theorem 9). A parametric Mehler covariance estimator is consistent under finite chaos order, and a convex-chord estimator is asymptotically conservative for any chaos order and any overstatement of the spatial range (Theorem 10). Block fractional sampling extends the collection calibration to dependent data (Theorem 11), under a linearization of the criterion tensor that we assume rather than prove; the selection procedure as a whole is therefore established for independent data only.

  • (iv) Evidence. Simulations generate Poisson incidence trajectories from a Gaussian-process (GP)–OU latent field on lattices and on U.S. state geography. An analysis of state syphilis surveillance fits a Poisson GP–OU model, reports a mixed-model event study alongside two-way fixed effects, and compares dependence-aware and conventional intervals for the functional contrast.

Section 3.4 connects the criterion to recent high-dimensional multivariate theory [6, 7, 8, 9, 10] through a deterministic stability bound, without claiming a high-dimensional inferential result. Two distinctions are essential throughout. An outer confidence collection protects against excluding a sufficient set; it does not certify every retained set, and at fixed error probability it does not support an oracle post-selection analysis. And Gaussian covariate laws within exposure groups are incompatible with nontrivial confounding under a uniform lower propensity bound, so we use strict positivity for identification and weighted conditions for estimation. The functional results concern outcomes isometric to a closed convex subset of a separable Hilbert space, such as univariate Wasserstein distributions, and complement Fréchet regression and distributional causal inference [11, 12] and scalar adjustment-set efficiency results [13, 14, 15]. The dependence theory assumes increasing-domain sampling and no interference. All proofs are in Appendix § 10.

The causal framework is set out in § 2: unconfoundedness and a relative-overlap condition for general stochastic interventions, the average treatment effect and incremental-intervention effects for distribution-valued outcomes, and natural direct and indirect effects under sequential ignorability (§ 4.3). Every resulting estimator solves stacked estimating equations, so the dependent-data theory of Section 5 covers all of them. Section 8 compares the approach, in theory and in application, with the point-process framework of Papadogeorgou et al. [16] and Mukaigawara et al. [17], implemented in the R package geocausal.

Causal framework and identification

Potential outcomes, interventions and causal assumptions

Let O=(X,W,Y)O = (X, W, Y), where X∈RpX \in\mathbb{R}^p, W∈W={1,…,K}W \in\mathcal{W} = \{1, \ldots, K\}, and YY takes values in a Polish space Ω\Omega. The dimensions pp and KK are fixed. Each unit has potential outcomes Y(w)Y(w), w∈Ww \in\mathcal{W}. Throughout we maintain the stable unit treatment value assumption [18]: consistency, Y=Y(W)Y = Y(W), without hidden versions of the exposure, and no interference, so that a unit’s potential outcomes do not depend on the exposures of other units. Section 5 lets units be dependent but keeps no interference; Section 8 contrasts this with point-process designs in which spillover is itself the object of study.

Distribution-valued outcomes are averaged through an isometry ι\iota from Ω\Omega onto a closed convex subset C\mathcal{C} of a separable Hilbert space H\mathcal{H}; write Y~=ι(Y)\widetilde{Y} = \iota(Y) and Y~(w)=ι{Y(w)}\widetilde{Y}(w) = \iota\{Y(w)\}. If E∥Y~∥2<∞\mathbb{E}\lVert\widetilde{Y}\rVert^2 < \infty, then EY~∈C\mathbb{E}\widetilde{Y} \in\mathcal{C} corresponds to the Fréchet mean, because E∥Y~−a∥2=E∥Y~−EY~∥2+∥a−EY~∥2\mathbb{E}\lVert\widetilde{Y} - a\rVert^2 = \mathbb{E}\lVert\widetilde{Y} - \mathbb{E}\widetilde{Y}\rVert^2 + \lVert a - \mathbb{E}\widetilde{Y}\rVert^2. For one-dimensional Wasserstein outcomes, ι\iota is the quantile embedding in L2(0,1)L^2(0,1).

An intervention FF is a conditional law of the exposure given the covariates. Let ff denote the density of FF with respect to counting measure on W\mathcal{W}; we let f(w∣x)f(w \mid x) depend on xx only through xSx_S for a covariate set S⊆[p]S \subseteq[p]. Setting W=aW = a is the deterministic intervention fa(w∣x)=1(w=a)f_a(w \mid x) = \mathbf{1}(w = a); a stochastic intervention spreads mass over several exposures [19, 20, 16]. The primary estimand is the average treatment effect (ATE) of arm aa versus arm bb,

Δ=E{Y~(a)−Y~(b)}∈H,\Delta= \mathbb{E}\{\widetilde{Y}(a) - \widetilde{Y}(b)\} \in\mathcal{H},

which contrasts the deterministic interventions faf_a and fbf_b. For Wasserstein outcomes it is the difference between the quantile functions of the Wasserstein barycentres of the two potential-outcome distributions [12]. Section 2.2 places it among the stochastic-intervention effects. The observed assignment mechanism has density e(w∣xS)=P(W=w∣XS=xS)e(w \mid x_S) = P(W = w \mid X_S = x_S), the propensity score.

We begin by introducing two causal assumptions. We first consider the unconfoundedness assumption, which ensures that the treatment assignment is random given the observed covariates [16, 21]. Formally, let ff denote the density of intervention FF. The unconfoundedness assumption in our context is stated as follows.

Assumption 1 (Unconfoundedness). For every exposure ww with P{f(w∣XS)>0}>0P\{f(w \mid X_S) > 0\} > 0 and every v∈Wv \in\mathcal{W},

P{W=v∣XS,Y(w)}=e(v∣XS)almost surely;P\{W = v \mid X_S, Y(w)\} = e(v \mid X_S) \quad\text{almost surely};

equivalently, Y(w)⊥W∣XSY(w) \perp W \mid X_S.

The second assumption replaces the bounded relative overlap condition e(w∣x)≥c0f(w∣x)e(w \mid x) \ge c_0 f(w \mid x), c0>0c_0 > 0, of Papadogeorgou et al. [16] by a moment condition. Let mw(x)=E(Y~∣XS=x,W=w)m_w(x) = \mathbb{E}(\widetilde{Y} \mid X_S = x, W = w) and σw2(x)=E{∥Y~(w)−mw(x)∥2∣XS=x}\sigma_w^2(x) = \mathbb{E}\{\lVert\widetilde{Y}(w) - m_w(x)\rVert^2 \mid X_S = x\}, with the convention fσw2/e=0f\sigma_w^2/e = 0 where f(w∣x)=0f(w \mid x) = 0.

Assumption 2 (Relative overlap). (a) e(w∣XS)>0e(w \mid X_S) > 0 almost surely on the event {f(w∣XS)>0}\{f(w \mid X_S) > 0\}. (b) E{∑wf(w∣XS)2σw2(XS)/e(w∣XS)}<∞\mathbb{E}\{\sum_w f(w \mid X_S)^2\sigma_w^2(X_S)/e(w \mid X_S)\} < \infty.

Part (b) holds under bounded relative overlap whenever Eσw2(XS)<∞\mathbb{E}\sigma_w^2(X_S) < \infty, but it does not require the ratio f/ef/e to be bounded. This matters for the ATE: for the deterministic interventions faf_a and fbf_b, a bounded ratio f/ef/e means uniform overlap, e(a∣XS)≥c0e(a \mid X_S) \ge c_0 and e(b∣XS)≥c0e(b \mid X_S) \ge c_0, which the Gaussian covariate laws of Proposition 1 exclude unless W⊥XW \perp X. Part (b) instead requires only E{σa2(XS)/e(a∣XS)+σb2(XS)/e(b∣XS)}<∞\mathbb{E}\{\sigma_a^2(X_S)/e(a \mid X_S) + \sigma_b^2(X_S)/e(b \mid X_S)\} < \infty.

Causal estimands and their identification

For an intervention FF define the interventional mean

μ(F)=E{∑wf(w∣XS)Y~(w)}∈C,\mu(F) = \mathbb{E}\left\{\sum_{w} f(w \mid X_{\mathcal{S}}) \widetilde{Y}(w)\right\} \in\mathcal{C},

the Fréchet mean of the potential outcome Y(WF)Y(W^{F}) when WFW^{F} is drawn from F(⋅∣X)F(\cdot\mid X) independently of the potential outcomes given XX. Since μ(fa)=EY~(a)\mu(f_a) = \mathbb{E}\widetilde{Y}(a), the ATE of § 2.1 is the contrast Δ=μ(fa)−μ(fb)\Delta= \mu(f_a)-\mu(f_b) of two deterministic interventions. More generally, a stochastic-intervention effect is τ(F′,F′′)=μ(F′)−μ(F′′)\tau(F', F'') = \mu(F')-\mu(F''). For binary exposure with e(x)=e(1∣x)e(x) = e(1 \mid x), the incremental intervention FδF_\delta of Kennedy [20] multiplies each unit’s odds of exposure by δ>0\delta> 0,

fδ(1∣x)=δe(x)δe(x)+1−e(x),(1)f_\delta(1 \mid x) = \frac{\delta e(x)}{\delta e(x) + 1 - e(x)}, \tag*{(1)}

and τδ=μ(Fδ)−EY~\tau_\delta= \mu(F_\delta) - \mathbb{E}\widetilde{Y} compares it with the observed assignment, which is the case δ=1\delta= 1. Binary exposures are labelled W={0,1}\mathcal{W} = \{0,1\}. Mediation estimands are defined in § 4.3.

References

  1. [1]W. Luo, F. Qin, L. Zhu, An exhaustive selection of sufficient adjustment sets for causal inference, arXiv preprint arXiv:2408.09415 (2024).arxiv.org/abs/2408.09415
  2. [2]W. Sheng, X. Yin, Direction estimation in single-index models via distance covariance, Journal of Multivariate Analysis 122 (2013) 148–161.DOI
  3. [3]W. Sheng, X. Yin, Sufficient dimension reduction via distance covariance, Journal of Computational and Graphical Statistics 25 (2016) 91–104.DOI
  4. [4]H.-H. Huang, F. Yu, T. Zhang, Robust sufficient dimension reduction via α-distance covariance, Journal of Nonparametric Statistics 37 (2025) 796–811. doi:10.1080/10485252.2024.2313137.DOI
  5. [5]H.-H. Huang, F. Yu, K. Li, T. Zhang, Fréchet sufficient dimension reduction for metric space-valued data via distance covariance, Journal of Computational and Graphical Statistics 35 (2026) 1117–1130. doi:10.1080/10618600.2025.2602521.DOI
  6. [6]T. Umino, K. Yata, M. Aoshima, Automatic sparse estimation of the high-dimensional cross-covariance matrix, Journal of Multivariate Analysis 213 (2026) 105590. doi:10.1016/j.jmva.2025.105590.DOI
  7. [7]K. Egashira, K. Yata, M. Aoshima, Asymptotic properties of hierarchical clustering in high-dimensional settings, Journal of Multivariate Analysis 199 (2024) 105251. doi:10.1016/j.jmva.2023.105251.DOI
  8. [8]K. Egashira, K. Yata, M. Aoshima, Asymptotic properties of k-means and its bias correction for high-dimension, low-sample-size data, Journal of Multivariate Analysis 216 (2026) 105675. doi:10.1016/j.jmva.2026.105675.DOI
  9. [9]A. Ishii, K. Yata, M. Aoshima, Geometric classifiers for high-dimensional noisy data, Journal of Multivariate Analysis 188 (2022) 104850. doi:10.1016/j.jmva.2021.104850.DOI
  10. [10]T. Bando, T. Sei, K. Yata, Consistency of the objective general index in high-dimensional settings, Journal of Multivariate Analysis 189 (2022) 104938. doi:10.1016/j.jmva.2021.104938.
  11. [11]A. Petersen, H.-G. Müller, Fréchet regression for random objects with Euclidean predictors, The Annals of Statistics 47 (2019) 691–719.arxiv.org/abs/1608.03012
  12. [12]Z. Lin, D. Kong, L. Wang, Causal inference on distribution functions, Journal of the Royal Statistical Society Series B: Statistical Methodology 85 (2023) 378–398. doi:10.1093/jrsssb/qkad008.DOI
  13. [13]J. Hahn, Functional restriction and efficiency in causal inference, The Review of Economics and Statistics 86 (2004) 73–76.DOI
  14. [14]A. Rotnitzky, E. Smucler, Efficient adjustment sets for population average causal treatment effect estimation in graphical models, Journal of Machine Learning Research 21 (2020) 1–86.
  15. [15]L. Henckel, E. Perković, M. H. Maathuis, Graphical criteria for efficient total effect estimation via adjustment in causal linear models, Journal of the Royal Statistical Society Series B: Statistical Methodology 84 (2022) 579–599.DOI
  16. [16]G. Papadogeorgou, K. Imai, J. Lyall, F. Li, Causal inference with spatio-temporal data: estimating the effects of airstrikes on insurgent violence in Iraq, Journal of the Royal Statistical Society, Series B 84 (2022) 1969–1999. doi:10.1111/rssb.12548.DOI
  17. [17]M. Mukaigawara, K. Imai, J. Lyall, G. Papadogeorgou, Spatiotemporal causal inference with arbitrary spillover and carryover effects: airstrikes and insurgent violence in the Iraq war, arXiv preprint arXiv:2504.03464, 2025.arxiv.org/abs/2504.03464
  18. [18]D. B. Rubin, Comment on “randomization analysis of experimental data: the Fisher randomization test” by D. Basu, Journal of the American Statistical Association 75 (1980) 591–593.
  19. [19]I. Díaz Muñoz, M. van der Laan, Population intervention causal effects based on stochastic interventions, Biometrics 68 (2012) 541–549.DOI
  20. [20]E. H. Kennedy, Nonparametric causal effects based on incremental propensity score interventions, Journal of the American Statistical Association 114 (2019) 645–656.arxiv.org/abs/1704.00211
  21. [21]I. Bojinov, N. Shephard, Time series experiments and causal estimands: exact randomization tests and trading, Journal of the American Statistical Association 114 (2019) 1665–1682.
  22. [22]C. J. Paciorek, The importance of scale for spatial-confounding bias and precision of spatial regression estimators, Statistical Science 25 (2010) 107–125.DOI
  23. [23]B. J. Reich, S. Yang, Y. Guan, A. B. Giffin, M. J. Miller, A. G. Rappold, A review of spatial causal inference methods for environmental and epidemiological applications, International Statistical Review 89 (2021) 605–634. doi:10.1111/insr.12452.DOI
  24. [24]R. D. Cook, Regression Graphics, Wiley, New York, 1998.
  25. [25]G. J. Székely, M. L. Rizzo, Partial distance correlation with methods for dissimilarities, The Annals of Statistics 42 (2014) 2382–2412.arxiv.org/abs/1310.2926
  26. [26]D. N. Politis, J. P. Romano, Large sample confidence regions based on subsamples under minimal assumptions, The Annals of Statistics 22 (1994) 2031–2050.DOI
  27. [27]C. F. J. Wu, On the asymptotic properties of the jackknife histogram, The Annals of Statistics 18 (1990) 1438–1452.DOI
  28. [28]J. Hájek, Limiting distributions in simple random sampling from a finite population, Publications of the Mathematical Institute of the Hungarian Academy of Sciences 5 (1960) 361–374.
  29. [29]J. M. Robins, S. Greenland, Identifiability and exchangeability for direct and indirect effects, Epidemiology 3 (1992) 143–155.pubmed.ncbi.nlm.nih.gov/1576220
  30. [30]J. Pearl, Direct and indirect effects, in: Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence, Morgan Kaufmann, San Francisco, 2001, pp. 411–420.
  31. [31]K. Imai, L. Keele, T. Yamamoto, Identification, inference and sensitivity analysis for causal mediation effects, Statistical Science 25 (2010) 51–71.
  32. [32]E. J. Tchetgen Tchetgen, I. Shpitser, Semiparametric theory for causal mediation analysis: efficiency bounds, multiple robustness and sensitivity analysis, The Annals of Statistics 40 (2012) 1816–1845.DOI
  33. [33]V. Didelez, A. P. Dawid, S. Geneletti, Direct and indirect effects of sequential treatments, in: Proceedings of the Twenty-Second Conference on Uncertainty in Artificial Intelligence, AUAI Press, Arlington, VA, 2006, pp. 138–146.
  34. [34]T. J. VanderWeele, S. Vansteelandt, J. M. Robins, Effect decomposition in the presence of an exposure-induced mediator-outcome confounder, Epidemiology 25 (2014) 300–306.DOI
  35. [35]J. Besag, J. York, A. Mollié, Bayesian image restoration, with two applications in spatial statistics, Annals of the Institute of Statistical Mathematics 43 (1991) 1–20.DOI
  36. [36]H. Rue, S. Martino, N. Chopin, Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations, Journal of the Royal Statistical Society: Series B 71 (2009) 319–392.DOI
  37. [37]S. Banerjee, B. P. Carlin, A. E. Gelfand, Hierarchical Modeling and Analysis for Spatial Data, 2nd ed., CRC Press, Boca Raton, FL, 2014.
  38. [38]H. Wendland, Scattered Data Approximation, Cambridge University Press, Cambridge, 2005.DOI
  39. [39]R. C. Bradley, Basic properties of strong mixing conditions. A survey and some open questions, Probability Surveys 2 (2005) 107–144.DOI
  40. [40]N. Jenish, I. R. Prucha, Central limit theorems and uniform laws of large numbers for arrays of random fields, Journal of Econometrics 150 (2009) 86–98.pubmed.ncbi.nlm.nih.gov/20161289
  41. [41]A. N. Kolmogorov, Y. A. Rozanov, On strong mixing conditions for stationary Gaussian processes, Theory of Probability and Its Applications 5 (1960) 204–208.DOI
  42. [42]P. Csáki, J. Fischer, On the general notion of maximal correlation, Magyar Tudományos Akadémia Matematikai Kutató Intézetének Közleményei 8 (1963) 27–51.
  43. [43]H. H. Kelejian, I. R. Prucha, HAC estimation in a spatial framework, Journal of Econometrics 140 (2007) 131–154.DOI
  44. [44]W. Bryc, W. Smoleński, Moment conditions for almost sure convergence of weakly correlated random variables, Proceedings of the American Mathematical Society 119 (1993) 629–635.DOI
  45. [45]Centers for Disease Control and Prevention, NCHHSTP AtlasPlus: state and county surveillance data for sexually transmitted infections, National Center for HIV, Viral Hepatitis, STD, and TB Prevention. Export redistributed at https://github.com/KrisePeng/The-Modern-Rise-of-STIs-in-the-U.S. (accessed 21 October 2026), 2024.
  46. [46]A. Baker, B. Callaway, S. Cunningham, A. Goodman-Bacon, P. H. C. Sant’Anna, Difference-in-differences designs: a practitioner’s guide, Journal of Economic Literature 64 (2026) 498–557. Replication data at https://github.com/pedrohcgs/JEL-DiD.
  47. [47]L. Tierney, J. B. Kadane, Accurate approximations for posterior moments and marginal densities, Journal of the American Statistical Association 81 (1986) 82–86.DOI
  48. [48]L. Knorr-Held, Bayesian modelling of inseparable space-time variation in disease risk, Statistics in Medicine 19 (2000) 2555–2567.
  49. [49]J. S. Hodges, B. J. Reich, Adding spatially-correlated errors can mess up the fixed effect you love, The American Statistician 64 (2010) 325–334.DOI
  50. [50]L. Zhou, K. Imai, J. Lyall, G. Papadogeorgou, Estimating heterogeneous treatment effects for spatio-temporal causal inference: how economic assistance moderates the effects of airstrikes on insurgent violence, arXiv preprint arXiv:2412.15128, 2024.arxiv.org/abs/2412.15128
  51. [51]M. Mukaigawara, L. Zhou, G. Papadogeorgou, J. Lyall, K. Imai, geocausal: an R package for spatio-temporal causal inference, OSF Preprints, doi:10.31219/osf.io/5kc6f. Package available at https://github.com/mmukaigawara/geocausal, 2024.DOI
  52. [52]A. S. Kechris, Classical Descriptive Set Theory, volume 156 of Graduate Texts in Mathematics, Springer, New York, 1995.DOI
  53. [53]R. Lyons, Distance covariance in metric spaces, The Annals of Probability 41 (2013) 3284–3305.
  54. [54]D. Nolan, D. Pollard, U-processes: rates of convergence, The Annals of Statistics 15 (1987) 780–799.DOI
  55. [55]R. P. Sherman, Maximal inequalities for degenerate U-processes with applications to optimization estimators, The Annals of Statistics 22 (1994) 439–459.DOI
  56. [56]R. P. Sherman, The limiting distribution of the maximum rank correlation estimator, Econometrica 61 (1993) 123–137.DOI
  57. [57]D. Nualart, The Malliavin Calculus and Related Topics, 2nd ed., Springer, Berlin, 2006.
  58. [58]A. W. van der Vaart, Asymptotic Statistics, Cambridge University Press, Cambridge, 1998.
  59. [59]T. W. Anderson, The integral of a symmetric unimodal function over a symmetric convex set and some probability inequalities, Proceedings of the American Mathematical Society 6 (1955) 170–176.DOI

Paper details

vlnc:W_gju39b8k
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Claude, Codex

View statement
Looking for related papers…
Contents