[ { "assumptions": [ "AR(1)-style data-generating process", "Innovations are independent with finite variance" ], "authors": [ "D. A. Dickey", "W. A. Fuller" ], "citation": "S01", "claims": [ "Classical t-distributions are invalid under unit-root null" ], "conclusions": [ "Dedicated critical values are required for unit-root inference" ], "contributions": [ "Derivation of nonstandard null distributions for unit-root tests" ], "future_work": [ "Augmentation for serially correlated errors" ], "key_equations": [ "y_t = rho y_{t-1} + e_t", "Delta y_t = gamma y_{t-1} + e_t, gamma = rho-1" ], "limitations": [ "Baseline formulation assumes simple AR structure" ], "source_type": "paper", "summary": "Introduces the Dickey-Fuller unit-root framework and asymptotic distributions for testing nonstationarity; critical for checking whether lottery feature time series are regime-stable before bias inference.", "title": "Distribution of the Estimators for Autoregressive Time Series with a Unit Root", "url": "https://doi.org/10.1080/01621459.1979.10482531", "year": 1979 }, { "assumptions": [ "Lag augmentation captures short-run dependence" ], "authors": [ "S. E. Said", "D. A. Dickey" ], "citation": "S02", "claims": [ "Augmentation yields valid unit-root tests under broader dynamics" ], "conclusions": [ "Lagged-difference terms improve robustness" ], "contributions": [ "ADF-style regression framework" ], "future_work": [ "Data-driven lag selection criteria" ], "key_equations": [ "Delta y_t = alpha y_{t-1} + sum_{i=1}^p beta_i Delta y_{t-i} + e_t" ], "limitations": [ "Lag-order selection affects power" ], "source_type": "paper", "summary": "Extends Dickey-Fuller testing to higher-order ARMA error structures using augmentation; directly relevant for dependence checks in lottery-derived feature sequences.", "title": "Testing for Unit Roots in Autoregressive-Moving Average Models of Unknown Order", "url": "https://doi.org/10.1093/biomet/71.3.599", "year": 1984 }, { "assumptions": [ "Long-run variance estimator is consistent" ], "authors": [ "D. Kwiatkowski", "P. C. B. Phillips", "P. Schmidt", "Y. Shin" ], "citation": "S03", "claims": [ "Provides complementary evidence to unit-root-null tests" ], "conclusions": [ "Joint ADF/KPSS use improves diagnosis" ], "contributions": [ "Stationarity-null testing framework" ], "future_work": [ "Robust high-frequency and long-memory variants" ], "key_equations": [ "KPSS = T^{-2} sum_{t=1}^{T} S_t^2 / hat{sigma}^2", "S_t = sum_{i=1}^{t} hat{e}_i" ], "limitations": [ "Bandwidth selection influences finite-sample size" ], "source_type": "paper", "summary": "Introduces KPSS stationarity test, complementary to ADF. Useful for symmetric evidence: rejecting both nulls flags modeling misspecification or structural instability.", "title": "Testing the null hypothesis of stationarity against the alternative of a unit root", "url": "https://doi.org/10.1016/0304-4076(92)90104-Y", "year": 1992 }, { "assumptions": [ "Weak dependence and consistent long-run variance estimation" ], "authors": [ "P. C. B. Phillips", "P. Perron" ], "citation": "S04", "claims": [ "No lag augmentation required for correction" ], "conclusions": [ "Alternative to ADF under complex errors" ], "contributions": [ "PP test statistics robust to nuisance autocorrelation" ], "future_work": [ "Adaptive bandwidth rules" ], "key_equations": [ "Z_t = t_{gamma} - (hat{lambda}-hat{gamma}_0)/(2 hat{sigma}^2) * (sum y_{t-1}^2)^{1/2}", "Z_alpha = T(hat{rho}-1) - T^{-2}(hat{lambda}-hat{gamma}_0)/(2hat{sigma}^2)" ], "limitations": [ "Kernel/bandwidth choice affects performance" ], "source_type": "paper", "summary": "Introduces nonparametric corrections for serial correlation and heteroskedasticity in unit-root testing; relevant when lottery features have unknown error dynamics.", "title": "Testing for a Unit Root in Time Series Regression", "url": "https://doi.org/10.1093/biomet/75.2.335", "year": 1988 }, { "assumptions": [ "Model residuals are approximately white noise under null" ], "authors": [ "G. M. Ljung", "G. E. P. Box" ], "citation": "S05", "claims": [ "Improved small-sample behavior versus Box-Pierce" ], "conclusions": [ "Useful post-fit diagnostic for dependence" ], "contributions": [ "Finite-sample corrected omnibus autocorrelation test" ], "future_work": [ "Adaptive lag-window selection" ], "key_equations": [ "Q = n(n+2) sum_{k=1}^{h} hat{rho}_k^2/(n-k)" ], "limitations": [ "Power depends on horizon h" ], "source_type": "paper", "summary": "Defines Ljung-Box Q-statistic for residual autocorrelation; central for testing independence assumptions in draw-level feature models.", "title": "On a Measure of Lack of Fit in Time Series Models", "url": "https://doi.org/10.1093/biomet/65.2.297", "year": 1978 }, { "assumptions": [ "Large-sample approximation" ], "authors": [ "G. E. P. Box", "D. A. Pierce" ], "citation": "S06", "claims": [ "Enables compact model adequacy testing" ], "conclusions": [ "Residual whiteness can be tested jointly" ], "contributions": [ "Omnibus check for residual autocorrelation" ], "future_work": [ "Small-sample corrections" ], "key_equations": [ "Q_{BP} = n sum_{k=1}^{h} hat{rho}_k^2" ], "limitations": [ "Finite-sample calibration weaker than Ljung-Box" ], "source_type": "paper", "summary": "Establishes Box-Pierce residual autocorrelation test, a precursor to Ljung-Box; useful as baseline independence diagnostic.", "title": "Distribution of Residual Autocorrelations in Autoregressive-Integrated Moving Average Time Series Models", "url": "https://doi.org/10.1080/01621459.1970.10481180", "year": 1970 }, { "assumptions": [ "IID null after prewhitening/modeling" ], "authors": [ "W. A. Brock", "W. D. Dechert", "J. A. Scheinkman", "B. LeBaron" ], "citation": "S07", "claims": [ "Can reject IID when linear tests do not" ], "conclusions": [ "Useful secondary independence screen" ], "contributions": [ "Nonlinear dependence diagnostic" ], "future_work": [ "Robust parameter selection strategies" ], "key_equations": [ "W_{m,epsilon} = sqrt{N} (C_m(epsilon)-C_1(epsilon)^m)/hat{sigma}_{m,epsilon}" ], "limitations": [ "Sensitive to embedding dimension and epsilon" ], "source_type": "paper", "summary": "BDS test detects nonlinear dependence missed by linear autocorrelation tests; relevant for checking subtle mechanism-driven structure in lottery features.", "title": "A test for independence based on the correlation dimension", "url": "https://doi.org/10.1080/07474939608800353", "year": 1996 }, { "assumptions": [ "Reference in-control distribution specified" ], "authors": [ "E. S. Page" ], "citation": "S08", "claims": [ "CUSUM is sensitive to small persistent shifts" ], "conclusions": [ "Suitable for local regime-break surveillance" ], "contributions": [ "Sequential mean-shift detection" ], "future_work": [ "Multivariate and nonparametric CUSUM" ], "key_equations": [ "S_t = max(0, S_{t-1} + x_t - k)", "Signal if S_t > h" ], "limitations": [ "Requires tuning k and h" ], "source_type": "paper", "summary": "Introduces CUSUM sequential change detection; supports rolling/local anomaly monitoring around operational lottery transitions.", "title": "Continuous Inspection Schemes", "url": "https://doi.org/10.1093/biomet/41.1-2.100", "year": 1954 }, { "assumptions": [ "Recursive residual framework" ], "authors": [ "R. L. Brown", "J. Durbin", "J. M. Evans" ], "citation": "S09", "claims": [ "Detects gradual and abrupt instability" ], "conclusions": [ "Essential for regime-consistent inference" ], "contributions": [ "Operational structural-stability diagnostics" ], "future_work": [ "Break-date estimators with confidence sets" ], "key_equations": [ "W_t = sum_{i=k+1}^{t} w_i / hat{sigma}", "CUSUMSQ_t = sum w_i^2 / sum w_i^2" ], "limitations": [ "Power varies by break type" ], "source_type": "paper", "summary": "Develops CUSUM and CUSUMSQ stability diagnostics for regression coefficients; relevant to testing whether estimated lottery effects persist across eras.", "title": "Techniques for Testing the Constancy of Regression Relationships Over Time", "url": "https://www.jstor.org/stable/2984889", "year": 1975 }, { "assumptions": [ "Piecewise linear model with finite number of breaks" ], "authors": [ "J. Bai", "P. Perron" ], "citation": "S10", "claims": [ "Consistent estimation of multiple breakpoints" ], "conclusions": [ "Framework supports regime segmentation" ], "contributions": [ "Break-date estimation and break-count testing" ], "future_work": [ "Critical-value refinements and software" ], "key_equations": [ "y_t = x_t' beta_j + u_t for t=T_{j-1}+1,...,T_j", "hat{T} = argmin_{T_1,...,T_m} sum_j sum_t u_t^2" ], "limitations": [ "Computational cost increases with candidate breaks" ], "source_type": "paper", "summary": "Provides estimation/testing for multiple structural breaks with unknown dates; directly applicable to rule/machine transition analysis.", "title": "Estimating and Testing Linear Models with Multiple Structural Changes", "url": "https://doi.org/10.2307/2998540", "year": 1998 }, { "assumptions": [ "Segment-wise parametric structure" ], "authors": [ "J. Bai", "P. Perron" ], "citation": "S11", "claims": [ "Feasible applied workflow for multiple breaks" ], "conclusions": [ "Supports data-driven break selection" ], "contributions": [ "Algorithms and tests for break detection" ], "future_work": [ "Finite-sample critical-value improvements" ], "key_equations": [ "UDmax = max_{m in [1,M]} supF(m|0)", "LWZ(m)=log(hat{sigma}_m^2)+k_m*c_T/T" ], "limitations": [ "Requires trimming parameter choices" ], "source_type": "paper", "summary": "Expands practical computation and inference for multiple-break models; useful for robust era splitting prior to hypothesis tests.", "title": "Computation and analysis of multiple structural change models", "url": "https://doi.org/10.1002/jae.659", "year": 2003 }, { "assumptions": [ "Additive segment cost and penalty" ], "authors": [ "R. Killick", "P. Fearnhead", "I. A. Eckley" ], "citation": "S12", "claims": [ "Substantially faster than naive exact dynamic programming" ], "conclusions": [ "Scales well for rolling diagnostics" ], "contributions": [ "Exact yet fast changepoint algorithm" ], "future_work": [ "Model-specific penalties and online variants" ], "key_equations": [ "F(t)=min_{tau= F(s)" ], "limitations": [ "Penalty calibration controls sensitivity" ], "source_type": "paper", "summary": "Introduces PELT algorithm for exact penalized changepoint detection in near-linear time; crucial for long lottery archives.", "title": "Optimal Detection of Changepoints With a Linear Computational Cost", "url": "https://doi.org/10.1080/01621459.2012.737745", "year": 2012 }, { "assumptions": [ "Segment-wise parametric cost models" ], "authors": [ "R. Killick", "I. A. Eckley" ], "citation": "S13", "claims": [ "Accessible tooling improves applied changepoint practice" ], "conclusions": [ "Software-backed reproducibility is feasible" ], "contributions": [ "Reference implementation for multiple CPD algorithms" ], "future_work": [ "Extended models and diagnostics" ], "key_equations": [ "min_{m,\tau} \\{sum_{j=0}^{m} C(y_{\tau_j+1:\tau_{j+1}}) + beta f(m)\\}" ], "limitations": [ "Method performance remains model-dependent" ], "source_type": "paper", "summary": "Operationalizes changepoint methods (including PELT) in reproducible software; relevant for transparent experiment pipelines.", "title": "changepoint: An R Package for Changepoint Analysis", "url": "https://doi.org/10.18637/jss.v058.i03", "year": 2014 }, { "assumptions": [ "Piecewise stationarity/parametric segments depending on method" ], "authors": [ "C. Truong", "L. Oudre", "N. Vayatis" ], "citation": "S14", "claims": [ "No single CPD method dominates all settings" ], "conclusions": [ "Method choice should track signal/noise and cost model" ], "contributions": [ "Taxonomy of exact/approximate CPD algorithms" ], "future_work": [ "Benchmark standardization" ], "key_equations": [ "hat{\\mathcal{T}}=argmin_{\\mathcal{T}} V(\\mathcal{T},y)+pen(\\mathcal{T})" ], "limitations": [ "Review scope excludes some online methods" ], "source_type": "paper", "summary": "Systematic review of offline CPD methods and tradeoffs; helps choose robust break-detection families for lottery-bias studies.", "title": "Selective review of offline change point detection methods", "url": "https://doi.org/10.1016/j.sigpro.2019.107299", "year": 2020 }, { "assumptions": [ "Independent or positively dependent tests for exact control" ], "authors": [ "Y. Benjamini", "Y. Hochberg" ], "citation": "S15", "claims": [ "Higher power than FWER controls in large-scale testing" ], "conclusions": [ "Suitable default correction for broad scans" ], "contributions": [ "FDR concept and practical step-up algorithm" ], "future_work": [ "Dependence-robust variants" ], "key_equations": [ "Find k=max\\{i: p_(i) <= (i/m)q\\}; reject H_(1),...,H_(k)" ], "limitations": [ "Under arbitrary dependence can be anti-conservative" ], "source_type": "paper", "summary": "Introduces BH procedure controlling FDR, foundational for scanning large combinatorial feature spaces without uncontrolled false positives.", "title": "Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing", "url": "https://www.jstor.org/stable/2346101", "year": 1995 }, { "assumptions": [ "Arbitrary dependency structures" ], "authors": [ "Y. Benjamini", "D. Yekutieli" ], "citation": "S16", "claims": [ "Controls FDR under broad dependence assumptions" ], "conclusions": [ "Useful conservative sensitivity check" ], "contributions": [ "Dependence-robust FDR control" ], "future_work": [ "Adaptive dependence-aware procedures" ], "key_equations": [ "q' = q / sum_{i=1}^{m}(1/i)", "Reject if p_(i) <= (i/m) q'" ], "limitations": [ "Can be overly conservative" ], "source_type": "paper", "summary": "Extends FDR control to general dependence (BY correction), providing conservative safeguards when lottery features are correlated.", "title": "The control of the false discovery rate in multiple testing under dependency", "url": "https://doi.org/10.1214/aos/1013699998", "year": 2001 }, { "assumptions": [ "Valid p-values; estimable proportion of true nulls" ], "authors": [ "J. D. Storey" ], "citation": "S17", "claims": [ "Can improve power relative to fixed-threshold BH" ], "conclusions": [ "Supports graded evidence ranking" ], "contributions": [ "Practical q-value framework" ], "future_work": [ "Robust pi0 estimation under dependence" ], "key_equations": [ "q(p)=min_{t>=p} \\pi_0 m t / R(t)" ], "limitations": [ "pi0 estimation sensitivity" ], "source_type": "paper", "summary": "Introduces q-values and direct FDR estimation via pi0, enabling more powerful large-scale inference pipelines.", "title": "A direct approach to false discovery rates", "url": "https://doi.org/10.1111/1467-9868.00346", "year": 2002 }, { "assumptions": [ "Large-scale hypothesis setting with valid p-values" ], "authors": [ "J. D. Storey", "R. Tibshirani" ], "citation": "S18", "claims": [ "q-values provide interpretable false-discovery calibration" ], "conclusions": [ "Encourages full-spectrum evidence reporting" ], "contributions": [ "Applied framework for q-value reporting" ], "future_work": [ "Dependence-aware q-value extensions" ], "key_equations": [ "pFDR = E(V/R | R>0)", "q_i = min_{t >= p_i} pFDR(t)" ], "limitations": [ "Dependence can affect calibration" ], "source_type": "paper", "summary": "Demonstrates q-value methodology at genome scale; transferable to lottery feature scans where hypotheses are numerous.", "title": "Statistical significance for genomewide studies", "url": "https://doi.org/10.1073/pnas.1530509100", "year": 2003 }, { "assumptions": [ "Mixture model for test statistics" ], "authors": [ "B. Efron" ], "citation": "S19", "claims": [ "Theoretical null can miscalibrate in practice" ], "conclusions": [ "Null-model diagnostics are necessary" ], "contributions": [ "Empirical-null perspective for large-scale testing" ], "future_work": [ "Flexible empirical Bayes null estimation" ], "key_equations": [ "fdr(z)=p_0 f_0(z)/f(z)" ], "limitations": [ "Modeling assumptions for z-density" ], "source_type": "paper", "summary": "Analyzes empirical-null choices in massive testing, useful when lottery-derived statistics show calibration drift from theoretical nulls.", "title": "Large-Scale Simultaneous Hypothesis Testing: The Choice of a Null Hypothesis", "url": "https://doi.org/10.1198/016214504000000089", "year": 2004 }, { "assumptions": [ "Statistical tests are screening tools, not proof of cryptographic security" ], "authors": [ "L. E. Bassham", "A. Rukhin", "J. Soto", "et al." ], "citation": "S20", "claims": [ "No finite battery can certify true randomness" ], "conclusions": [ "Use batteries as first-stage diagnostics" ], "contributions": [ "Standardized multi-test randomness assessment" ], "future_work": [ "NIST revision for updated methods" ], "key_equations": [ "P-value = igamc(a, x)", "Uniformity test via chi^2 across P-value bins" ], "limitations": [ "Not sufficient for security claims" ], "source_type": "paper", "summary": "Reference battery of randomness tests; provides a transferable template for binary-sequence diagnostics derived from lottery encodings.", "title": "A Statistical Test Suite for Random and Pseudorandom Number Generators for Cryptographic Applications", "url": "https://doi.org/10.6028/NIST.SP.800-22r1a", "year": 2010 }, { "assumptions": [ "Generator outputs mapped to U(0,1) under null" ], "authors": [ "P. L'Ecuyer", "R. Simard" ], "citation": "S21", "claims": [ "Different generators fail different deep tests" ], "conclusions": [ "Battery diversity matters" ], "contributions": [ "Comprehensive RNG testing library and methodology" ], "future_work": [ "Additional tests and modern implementations" ], "key_equations": [ "U_i = F(X_i) for uniformization under null" ], "limitations": [ "Computationally expensive large batteries" ], "source_type": "paper", "summary": "Introduces TestU01 batteries (SmallCrush/Crush/BigCrush); useful as high-power alternatives when testing transformed lottery sequences.", "title": "TestU01: A C Library for Empirical Testing of Random Number Generators", "url": "https://doi.org/10.1145/1268776.1268777", "year": 2007 }, { "assumptions": [ "Finite-state linear recurrence over F2" ], "authors": [ "M. Matsumoto", "T. Nishimura" ], "citation": "S22", "claims": [ "Improved quality/speed tradeoff vs earlier generators" ], "conclusions": [ "Suitable baseline for simulation studies" ], "contributions": [ "Long-period, high-equidistribution PRNG" ], "future_work": [ "Variants with better equidistribution/seeding" ], "key_equations": [ "x_{k+n}=x_{k+m} \\oplus ((x_k^u | x_{k+1}^l)A)", "y = x \\oplus (x>>u); y = y \\oplus ((y<