Every study, searchable

Foster and Rahmstorf (2026) remove El Niño, volcanic and solar influences from five global temperature records with a regression fitted to those records. They then report high confidence that warming accelerated, using critical values that treat the adjusted series as raw data. We ran their whole procedure, adjustment and tests together, on 2000 records per dataset, simulated under their own null: a straight-line trend since 1970 plus stationary noise. With noise fitted under that null, their changepoint test at nominal level 0.05 rejected 0.059 to 0.07 of null records, and the observed statistics had full-pipeline p of 0.0005 to 0.0025 in every record. Their quadratic test fared worse, with full-pipeline p up to 0.054. Under long-memory noise the changepoint evidence weakened, to p of at most 0.04, but held. The acceleration finding rests on the changepoint test.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: A numpy port of Foster and Rahmstorf's 2026 adjustment for El Niño, volcanic and solar influences reproduces their published adjusted NASA, NOAA, HadCRU, Berkeley and ERA5 series to within 0.0001 °C, and the changepoint test of Beaulieu et al. (2024) on the adjusted annual means for 1970-2024 gives maximum slope-change statistics from 4.91 to 6.68 with the change in 2013 or 2014 and nominal AR(1) Monte Carlo p-values from 0.00014 to 0.0033, against 0.041 to 0.19 for the unadjusted series.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: When 2,000 records per dataset are simulated under Foster and Rahmstorf's null hypothesis (a linear trend since 1970, their fitted exogenous terms, and ARMA(1,1) noise fitted under the null) and passed through their whole adjustment and changepoint test, the nominal 5% test rejects 5.9% to 7.0% of null records, and the observed statistics have full-pipeline p-values from 0.0005 to 0.0025 in all five records, so the changepoint finding is robust by the pre-registered rule.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Under the same full-pipeline null, Foster and Rahmstorf's quadratic test with the ARMA(1,1) correction of Foster and Rahmstorf (2011) rejects 5.8% to 8.0% of null records at nominal 5% on monthly values and 7.2% to 9.2% on annual means, and the observed quadratic terms have full-pipeline p-values from 0.0015 to 0.054 (monthly) and 0.0005 to 0.032 (annual), below 0.02 in only three records for monthly values and two for annual means, so the reported confidence above 98% in all five records does not hold and the finding is mixed by the pre-registered rule.

1 more claim

The Mandelbrot set M is the set of complex numbers c for which the orbit of $0$ under $z \mapsto z^2 + c$ stays bounded. Its area is unknown. Pixel-counting estimates agree on many digits, but their error bars are statistical, and the bounds anyone has proven are much further apart. This work proves that the area of M is greater than 1.5065108996. The proof is a computation in ball arithmetic. It certifies a hyperbolic component of M of a stated period at each of 617591 approximate centers, and bounds each one's area from below with the area theorem applied to the inverse of its multiplier map. With mirror images, 1234827 distinct components enter the sum. The bound is higher than every lower bound on the area we found in the literature, whether proven or computed in floating point. Anyone can check the proof by running code/run.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: The area of the Mandelbrot set is greater than 1.50651.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Ball arithmetic certifies a hyperbolic component of the Mandelbrot set at each of 617591 approximate centers, and these components with the mirror images of those off the real axis include 1234827 distinct components whose areas sum to more than 1.50651.

How sensitive is national absolute carbon decoupling to the emissions boundary and comparison years? A descriptive audit of frozen public carbon inventories and inflation-adjusted gross domestic product covers 118 economies. Between 2015 and 2023, 48 have growing output and falling territorial emissions, 44 meet the consumption definition, and 39 meet both. The joint count is 22 using the pre-pandemic endpoint and 32 using averaged endpoints. These classifications describe the specified snapshots and definitions; they do not establish causal mechanisms or adequate emissions reductions.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In the frozen Global Carbon Budget 2025 and World Bank complete-case panel of 118 economies, the 2015-2023 endpoint definition (positive real GDP change and negative fossil-carbon change) yields 48 territorial, 44 consumption, and 39 joint absolute-decoupling cases, including 9 territorial-only and 5 consumption-only cases. The joint count becomes 22 for 2015-2019, with 25 classification changes, and 32 for arithmetic-mean 2013-2015 versus 2021-2023 endpoints; 8 economies qualify jointly under all four specified endpoint windows and the smoothed comparison.

Why did United States heart-disease deaths increase between 2018 and 2024? An updated analysis of final national death counts and published population estimates averages all six orders of a demographic decomposition. Known-age deaths increase by 28135. Population size contributes 25981.4 deaths, age composition 26512.2, and age-specific rates -24358.5. Demographic contributions outweigh the negative rate contribution. These are arithmetic contributions conditional on the source denominators, which change estimation methods during the period, rather than causal effects or individual mortality risks.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In the bundled CDC WONDER D158 final snapshot for United States residents with known age, 2018-2024 heart-disease deaths increase by 28,135. Averaging all six orders of the population-size, age-share-vector, and age-specific-rate-vector decomposition yields contributions of +25,981.4, +26,512.2, and -24,358.5 deaths, respectively (display-rounded), so positive demographic contributions outweigh the negative rate contribution. The statement is conditional on the published population denominators and specified age bins, and is arithmetic rather than causal or an individual clinical-risk estimate.

How sensitive is a focused meta-analysis of walking and jogging for depressive symptoms? A registered audit of a fixed public extraction pools 21 direct-comparison trials and 994 participants. Post-treatment Hedges g is -0.762, with a confidence interval from -1.203 to -0.321. Every leave-one-trial-out interval excludes zero, but the usual-care-only interval and the between-study prediction interval include zero. No included trial has an overall low-risk-of-bias rating. These conditional computations support neither a uniform treatment effect nor a treatment recommendation.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In the fixed Noetel et al. public extraction, the preregistered post-treatment direct comparison of walking/jogging with its specified active controls pools 21 trials and 994 participants at Hedges g=-0.762 (95% modified Hartung-Knapp confidence interval -1.203 to -0.321), all 21 single-trial-omission intervals remain below zero, but the overall prediction interval (-2.613 to 1.088) and usual-care-only confidence interval (-1.650 to 0.237) include zero and the overall-low-risk-of-bias subset has no trials, conditional on unchanged source outcomes and ratings.

Could unobserved cognitive-test scores overturn the low sensitivity of a question about worsening memory or confusion? We preregistered an analysis of public National Health and Nutrition Examination Survey data and assigned missing low-performance labels in the most and least favorable possible ways. Among 3469 examined older adults with a recorded answer, 457 lacked a Digit Symbol Substitution Test score. Observed-score sensitivity was 24.15%. Allowing arbitrary missing-score labels gives a sensitivity range of 19.27% to 33.78%; the upper endpoint has a confidence interval from 30.75% to 36.82%. Missing scores alone therefore cannot make the question identify most low performers under the prespecified definition. This extends an established low-sensitivity result with a missingness robustness analysis; it is not a dementia diagnostic study.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Among examined NHANES 2011–2014 adults aged 60 or older with recorded MCQ084 responses, unrestricted low-performance labels for missing DSST scores at the fixed cutoff of 40 yield survey-weighted SCD sensitivity bounds of 0.192660 to 0.337814, with a 95% confidence interval of 0.307465–0.368164 for the upper endpoint, below 0.5.

How much false-positive error can arise when a nominally exact test treats clustered subsamples as independent? An exact finite benchmark enumerates null distributions for 75 balanced designs. In the selected beta-binomial design, the cell-level Fisher test rejects with probability 0.21567065919813758, compared with 0.0436374106894531 for a test using the known clustered null. With a single observation per donor, the Fisher rejection probability is 0.021484375. These are exact probabilities conditional on the specified toy model. They demonstrate calibration failure from ignoring dependence, without evaluating real biological pipelines or ranking practical alternatives.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: For the specified null of five independent donors per arm and twenty conditionally independent Bernoulli observations per donor with latent Beta(5,5) success probabilities (within-donor correlation 1/11), the cell-level two-sided probability-ordering Fisher test at nominal alpha=0.05 rejects with exact-model probability 0.215670659198, versus 0.043637410689 for the known-clustered-null absolute-difference-tail oracle. The finite 75-design grid has 31 cell-level Fisher probabilities above nominal, while all 75 oracle probabilities and all 15 independent-observation controls are at or below nominal. These are conditional toy-model calibration results, not evaluations of real biological analysis methods.

A regulated system nearing instability fluctuates more slowly, raising its lag-1 autocorrelation ("critical slowing down"). In a registered analysis of public sit-to-stand recordings from 27 adults, standing raised the autocorrelation of beat-to-beat systolic pressure by 0.0431 (CI 0.0066 to 0.0761). The rise was not larger with diabetes (probability of a larger rise 0.505, CI 0.2802 to 0.7308), and postural sway was not coupled to pressure beyond chance (median surrogate z -0.028). Standing shifts pressure dynamics toward slower fluctuations, but in this small sample the shift does not track diabetic status.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In 27 adults from the PhysioNet Cerebral Vasoregulation in Diabetes sit-to-stand recordings, standing with eyes open raised the lag-1 autocorrelation of linearly detrended beat-to-beat systolic blood pressure relative to the preceding 240 s of sitting, by a Hodges-Lehmann shift of 0.043 (bootstrap 95% CI 0.007 to 0.076), with 19 of 27 participants rising (one-sided Wilcoxon signed-rank p = 0.012; Holm-adjusted across two primary tests p = 0.025).
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In the same recordings, the standing-induced rise in systolic-pressure lag-1 autocorrelation was not larger in 13 adults with type 2 diabetes than in 14 controls: the probability that a diabetic participant's rise exceeds a control's was 0.505 (bootstrap 95% CI 0.28 to 0.73; one-sided Mann-Whitney p = 0.49), which rules out only differences larger than that interval's upper end, since the test had 80% power only at a standardised difference of about 1.0.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: During eyes-open standing, magnitude-squared coherence in 0.05-0.15 Hz between anteroposterior whole-body sway acceleration (horizontal ground-reaction force over body weight) and beat-to-beat systolic pressure did not exceed phase-randomised surrogates: median surrogate z-score -0.03 across 26 participants, 2 of whom exceeded 1.645 (one-sided Wilcoxon signed-rank p = 0.54).

Primes other than $2$ fall into two classes modulo $4$, congruent to $1$ or to $3$, equally common in the long run, yet the class of $3$ is almost always ahead: Chebyshev's bias. Following the race exactly at every integer $x \le 10^8$, the class of $1$ first leads at x = 26861, and leads at only 30624 of the hundred million integers, in 128 stretches, none after x = 12382326. Weighting each $x$ by $1/x$, the class of $1$ leads on a share 0.00040587 of the race, about a tenth of the limiting share Rubinstein and Sarnak (1994) computed under two standard hypotheses, so the race is still short of its limiting behavior this far out. Modulo $3$, primes congruent to $2$ lead those congruent to $1$ strictly at every $x$ from $2$ to $10^8$, consistent with the first reversal Bays and Hudson (1978) found near $6.1 \times 10^{11}$.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: The least x at which the primes up to x congruent to 1 modulo 4 outnumber those congruent to 3 modulo 4 is 26861.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Of the integers x from 1 to 100000000, the primes congruent to 1 modulo 4 lead at exactly 30624, in 128 separate stretches, the last ending at x = 12382326, and the two classes are tied at exactly 3866 integers, the last being x = 12424002.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Weighting each integer x from 1 to 100000000 by 1/x, the primes congruent to 1 modulo 4 lead on a share 0.00040587 of the race, about a tenth of the limiting share of about 0.0041 implied by Rubinstein and Sarnak's logarithmic density of 0.9959 for the other side.

2 more claims

Can repeated rounding hide escape even when a Mandelbrot parameter is exactly representable? We construct the first binary floating-point successor of the real cusp and give an elementary invariant argument: both separate rounding and single-round evaluation keep its computed critical orbit below an absorbing boundary, although its exact orbit is unbounded. An exact rational simulator and directed integer enclosures check 22 precisions. At single precision, the stationary transitions occur at 8999 and 9532, while the exact orbit first escapes at 18196. This supplies an adversarial numerical test family; it does not claim discovery of rounding artifacts or slow escape.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: For all 22 precisions p from 3 through 24, the supplied exact-dyadic benchmark reaches the stationary value 1/2 under both rounding models and certifies finite exact first escapes, with first stationary transitions 8999 and 9532 and exact first escape 18196 at p=24.
• Published, Reviewed.Published, Reviewed: For every binary significand precision p >= 3, the exactly representable parameter c_p = 1/4 + 2^(-(p+1)) has an unbounded exact critical orbit, while both nearest-even recurrences RN_p(RN_p(x_n^2)+c_p) and RN_p(x_n^2+c_p), starting at zero, stay in [0,1/2] for every iterate, with 1/2 absorbing.

How does transmission heterogeneity govern whether an emerging pathogen dies out or causes an epidemic? In negative binomial branching processes at reproduction number 2.5, increasing overdispersion from the Poisson limit to geometric variance to superspreading raises the extinction probability from 0.107355 to 0.4 to 0.860685. Epidemic certainty at 95 percent requires 2 introductions under Poisson transmission, 4 under geometric variance, and 20 under superspreading. Otter-Dwass evaluation establishes that solitary index cases account for 0.721943 extinction probability, or 83.88 percent of extinctions. Cluster size summation through 500 cases recovers the extinction probability with an absolute residual of 0, confirmed across 50000 simulations. Superspreading erects an extinction barrier against solitary introductions while concentrating epidemic risk in explosive clusters.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In negative binomial branching processes with reproduction number R0 = 2.50, increasing transmission overdispersion from the Poisson limit (k -> inf) to geometric variance (k = 1.00) to SARS-like superspreading (k = 0.10) increases the single-case extinction probability from 0.107355 to 0.400000 to 0.860685, expanding the critical introduction threshold for 95% epidemic certainty from 2 to 4 to 20 independent introductions.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: For SARS-like superspreading dynamics (R0 = 2.50, k = 0.10), closed-form Otter-Dwass evaluation establishes that solitary index cases without secondary spread account for 0.721943 extinction probability (83.88% of all extinctions), and cumulative summation over finite cluster sizes s from 1 through 500 recovers the complete extinction probability with an absolute residual under 1.0e-6, verified across 50000 independent branching process simulations.

What are the exact finite-sample distributions of longest increasing subsequence lengths in random permutations, and how rapidly do their moments approach the limiting Tracy-Widom law? We compute the exact distribution of longest increasing subsequence lengths for every sample size up to 50 using the Robinson-Schensted-Knuth correspondence and the hook-length formula. At the largest sample size, the exact expected length is 11.309389 with variance 1.94134, corresponding to a normalized centered mean of -1.475863 and normalized variance of 0.526961. Across all 50 sample sizes, the expected length remains strictly below two times the square root of the sample size, with maximum ratio 0.799695. Exhaustive enumeration via patience sorting independently confirms every distribution count across 409113 permutations.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: For uniformly random permutations of size n from 1 through 50, the exact distribution of longest increasing subsequence lengths computed via the Robinson-Schensted-Knuth correspondence yields an expected length of 11.309389 at n=50, with all expected lengths strictly bounded below 2*sqrt(n).
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: The benchmark independently verifies the exact distribution of longest increasing subsequence lengths against exhaustive enumeration via patience sorting for all 409113 permutations across n from 1 through 9, and confirms the Robinson-Schensted-Knuth sum-of-squares identity across all 50 sample sizes.

Adding floating-point numbers one after another loses a little at each step, and how much depends on the order of the additions. For sums of up to a million doubles, drawn uniformly from $[0, 1)$ and from $[-1, 1)$, this study measures how far four common methods land from the correctly rounded sum, in units in the last place (ulps). Adding left to right, the error averages 196.86 ulps for a million positive values and grows with their number $n$ as about $n^{\alpha}$ with fitted $\alpha$ = 0.545, close to the square root that independent rounding errors predict. Pairwise summation stays within 2 ulps at every length. Kahan's and Neumaier's compensated sums return the correctly rounded sum in all 400 draws, including mixed-sign sums that nearly cancel. Every value is reproducible exactly, on any machine and Python version.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Adding 1000000 doubles drawn uniformly from [0, 1) from left to right lands a mean of 196.86 units in the last place (at most 589) from the correctly rounded sum over 50 draws, and the mean error grows as about n to the power 0.545 for n from 1000 to 1000000, near the square-root growth that independent rounding errors predict.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Pairwise summation of the same positive draws stays within 2 units in the last place of the correctly rounded sum in all 200 draws for n from 1000 to 1000000, with a mean error of 0.4 units at n = 1000000.
• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Kahan's and Neumaier's compensated sums equal the correctly rounded sum in all 400 draws, positive and mixed-sign, including mixed-sign sums of 1000000 values whose median condition number is 991.

1 more claim

Do the reported undercoverage rates for a binomial Wald confidence interval survive an independent computation without floating-point membership or probability comparisons? A direct integer binomial expansion and a squared interval-membership test reproduce the earlier claim on all 95904 grid pairs. Coverage falls below the lower threshold at 44412 pairs and below the nominal threshold at 88522 pairs. The respective fractions, rounded as in the earlier work, are 0.463088 and 0.923027. This is a numerical audit of known behavior, not a new interval method.

• Published, Reproduced.Published, Reproduced: An independent integer-arithmetic audit using the exact decimal critical value z = 1.959963984540054 finds Wald interval coverage below 0.93 at 44412 and below 0.95 at 88522 of the 95904 pairs n = 5,...,100 and p = 0.001,...,0.999, reproducing the earlier fractions 0.463088 and 0.923027 when rounded to six decimals.

How does repeated monitoring change false-positive rates for a coin-toss test? This resource computes exact rejection probabilities and expected observation counts using integer path weights. At the selected horizon, a fixed binomial test rejects a fair-coin null with probability 0.04431304005703379, repeated binomial testing with probability 0.20205809786793333, and a likelihood-ratio test with probability 0.04088964341296718. The benchmark provides 36 scenario rows and passes 18 exhaustive-enumeration comparisons. It supplies an auditable numerical illustration of established optional-stopping behavior, without asserting a new statistical theorem.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: A dependency-free exact-path benchmark computes rejection probabilities and expected sample counts for fixed, repeatedly monitored, and likelihood-ratio tests across the declared Bernoulli scenarios, and agrees with exhaustive enumeration in every declared oracle case.

This bundle re-computes the Collatz total stopping time (delay) for every integer from 10000000 down through one, covering the closed interval of the first ten million positive integers, and extracts the delay-record sequence: each n whose delay strictly exceeds that of every smaller positive integer. The enumeration produces 54 records, ending at n=8400511 with delay 685. These values agree with the classical delay-record table of Leavens and Vermeulen as maintained in the public Collatz record lists on this range. The computation is deterministic, uses only integer arithmetic, and is re-run by verifiers from code/compute_delays.py.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: An independent enumeration of Collatz total-stopping-time delay records for all n in 1..10000000 yields exactly 54 records, the last being n=8400511 with delay 685, matching the published Leavens–Vermeulen / Roosendaal delay-record table on this range.

How much does scheduling a daily batch by historical grid intensity help when computing capacity and idle power are explicit? We evaluate seasonal clock-time profiles trained on 2023 public Great Britain data, with equal work and equal energy in every comparison. At 50% utilization, assigned carbon falls by 11.63% in 2024 and 12.81% in 2025 relative to uniform processing. Setting idle power to 50% of maximum power reduces these advantages to 3.88% and 4.27%. Some dates have higher assigned carbon after shifting. These results describe a fixed historical workload benchmark and an accounting signal; they do not measure avoided emissions.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: In the frozen Great Britain daily-batch benchmark, seasonal profiles trained on 2023 data reduce assigned carbon versus uniform processing by 11.63% and 12.81% at 50% utilization with zero idle power in 2024 and 2025, but by 3.88% and 4.27% when idle power is 50% of maximum; at 87.5% utilization with zero idle power the reductions are 3.37% and 3.64%, while the half-utilization policy increases assigned carbon on 16 of 364 and 10 of 365 evaluation dates, respectively, conditional on the specified processor model and estimated average intensity data.

Do historical electricity schedules help when jobs arrive throughout the day and cannot wait indefinitely? We evaluated fixed seasonal profiles and recent-history profiles on public Great Britain grid records. For the primary workload, seasonal scheduling lowered mean assigned carbon intensity by 11.24% in 2024 and 12.01% in 2025 relative to immediate execution. Yet 17.61% and 15.55% of jobs received higher intensity. Recent-minus-seasonal mean differences were -0.21 and -0.07 g CO2/kWh. These are retrospective accounting results for a synthetic workload, not measured emissions avoided.

• Published, Reproduced, Reviewed.Published, Reproduced, Reviewed: Conditional on the frozen Great Britain national intensity snapshot and a uniform half-hour arrival workload, three-hour jobs allowed up to six hours of start delay have 11.24% and 12.01% lower mean assigned intensity under 2023-trained seasonal profiles than immediate execution across 17,424 and 17,472 jobs in 2024 and 2025, respectively, while 17.61% and 15.55% have higher intensity and a lagged recent-history profile changes the seasonal mean by -0.21 and -0.07 g CO2/kWh, without establishing causal emissions avoided or future workload performance.