archive
Every paper Pith has read. Search by title, abstract, or pith.
883 papers in math.ST · page 7
-
Emergent AI abilities arise from infinite limit architecture
A Limit Theory of Foundation Models: A Mathematical Approach to Understanding Emergent Intelligence and Scaling Laws
-
Jackknife delivers conservative variance estimates under interference
Neyman Jackknife: Design-Based Variance Estimation for Causal Inference under Interference
-
Generator weights from triangular inversion characterize signed tail compatibility
A Geometric Witness Framework for Signed Multivariate Tail-Dependence Compatibility: Asymptotic Structure and Finite-Threshold Synthesis
-
The paper introduces a new metric on counting measures to define a Wasserstein distance…
Wasserstein convergence rates for empirical measures of point processes
-
Bayesian smoothing recovers sharp change-plane boundaries
Bayesian change-plane regression
-
Kac's walk on rotations mixes in n² log n steps
Kac's walk on rotation matrices mixes in $n^2 \log n$ steps
-
This paper introduces weak moment methods for statistical inference by representing…
Weak Moment Methods for Statistical Inference: with an Application to Robust Estimation
-
Preconditioner must stabilize faster than t^-(α+1)/2 to keep CLT
When Does Dynamic Preconditioning Preserve the Polyak-Ruppert CLT? A Stabilization Threshold
-
Diseased-case correlation raises rule-out ROC AUC
A theory of ROC analysis of rule-out and rule-in diagnostics with applications to mammography data
-
Rerandomization lowers variance of Kaplan-Meier survival estimators
Asymptotic theory of rerandomization for survival analysis
-
Conway-Maxwell multivariate Bernoulli spans full dependence spectrum
Conway--Maxwell multivariate Bernoulli distribution
-
Random and deterministic scan MCMC share spectral gap positivity
Solidarity of Spectral Gaps for Component-Wise Markov Chains
-
Spectral algorithms split into three risk regimes in high dimensions
Learning Curves and Benign Overfitting of Spectral Algorithms in Large Dimensions
-
Kendall adaptations test independence with missing high-dim data
Testing independence in the presence of missing data: high-dimensional case
-
Subanalytic push-forwards admit Sobolev embeddings into inner Lipschitz functions
Sobolev embedding theorem and subanalytic measures
-
Stability bounds keep generator errors from ruining output distributions
Statistical Analysis of Markovian Generative Modeling
-
Parity polytopes fix sharp bounds on random products
Sharp bounds for products of dependent random variables
-
Laplace-Stein test validates Pareto fits with competitive power
Laplace Transform driven Stein-type Goodness-of-fit Tests for Pareto Distribution
-
Adapted Wasserstein barycenters of Gaussians are Gaussian
Adapted Wasserstein Barycenters of Gaussian Processes
-
Monge map estimator from dual of sampled OT problem
Statistical Estimation of Monge Transport Maps via Brenier Potentials
-
Bandits optimize concave utilities via influence-function gradients
Concave Statistical Utility Maximization Bandits via Influence-Function Gradients
-
LS designs stay optimal for M-estimation asymptotically
Minimax Robust Designs for M-Estimated Models
-
Multicalibration needs Θ(ε^{-3}) samples for polynomial groups
The Sample Complexity of Multicalibration
-
Generalized Tweedie identity handles unequal variances in empirical Bayes
Nonparametric f-Modeling for Empirical Bayes Inference with Unequal and Unknown Variances
-
Mixture e-processes yield power-one tests for stochastic dominance
Betting on Bets: Anytime-Valid Tests for Stochastic Dominance
-
FPCA loses all signal once functional data crosses a roughness threshold
Does PCA Work for Rough Functional Data?
-
New test checks logistic propensity scores under nonignorable missing data
A goodness-of-fit test for the logistic propensity score model under nonignorable missing data
-
-
Weak moments define a central limit theorem for heavy-tailed laws
Distributional Statistical Models: Weak Moments, Cumulants, and a Central Limit Theorem
-
Framework defines moments for distributions without them
Distributional Statistical Models: Weak Moments, Cumulants, and a Central Limit Theorem
-
E-processes deliver power-one tests for monotonicity and unimodality
E-values and sequential power-one tests for monotonicity and unimodality
-
Doob transform yields consistent estimators for composite births
Likelihood-based inference for birth-death processes with composite birth mechanisms
-
Hilbert space approximation makes IMSE closed-form
Fast and Provably Accurate Sequential Designs using Hilbert Space Gaussian Processes
-
Bayesian methods estimate densities nonparametrically in three ways
Bayesian approaches to non- and semiparametric density estimation [with a rejoinder to my discussants]
-
Mixture betting yields ln ln n regret almost surely
Cover meets Robbins while Betting on Bounded Data: $\ln n$ Regret and Almost Sure $\ln\ln n$ Regret
-
Mixture betting yields ln ln n regret almost surely
Cover meets Robbins while Betting on Bounded Data: $\ln n$ Regret and Almost Sure $\ln\ln n$ Regret
-
Vertex misalignment impairs network changepoint detection only in joint cases
Vertex misalignment and changepoint localization in network time series
-
Weighted p-value ordering boosts power in Holm multiple testing
Weighted Holm Procedures: Theory, Properties, and Recommendations
-
Summed SNR from mixed hyperedges crosses detection threshold
Achieving the Kesten-Stigum bound in the non-uniform hypergraph stochastic block model
-
Compression reduces NPMLE cost to logarithmic in sample size
Fast computation and theoretical guarantees for the NPMLE in exponential family mixtures
-
DPP sampling improves Monte Carlo integration rates
On two ways to use determinantal point processes for Monte Carlo integration
-
Resource allocation matrix beats binary incidence for hypergraphs
Hypergraph Mining via Proximity Matrix
-
Nonconvex LASSO critical points match convex recovery rates
Sharp recovery and landscape guarantees for the nonconvex matrix LASSO
-
Asymptotic e-processes bound excursions up to finite horizon r_m
Asymptotic e-processes
-
Asymptotic e-processes bound excursions uniformly up to growing horizons
Asymptotic e-processes
-
Stochastic intervention finds optimal treatment distributions as options grow
Stochastic Intervention
-
Excluding neighbourhoods sets loss-based priors by local KL geometry
The General Formulation of Loss-Based Priors for Parameter Spaces
-
Clustering limits phase-transition between weak and strong communities
The interplay between network transitivity and community structure
-
Inhomogeneous noise can improve signal detection
BBP transition and the leading eigenvector of the spiked Wigner model with inhomogeneous noise
-
Anisotropy creates temporary teacher spike in gradient flow
Random Matrix Theory of Early-Stopped Gradient Flow: A Transient BBP Scenario