Systematic assessment of the Hubble tension via Bayesian jackknife testing 2026 Systematic assessment of the Hubble tension via Bayesian jackknife testing – A

Thomas Hughes, Michael J. Wilensky, Philip Bull · 2025

Read Systematic assessment of the Hubble tension via Bayesian jackknife testing 2026 Systematic assessment of the Hubble tension via Bayesian jackknife testing – A by Thomas Hughes, Michael J. Wilensky, Philip Bull (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-19341.txt.

Introduction

Thomas Hughes, Michael J. Wilensky, and Philip Bull probe the Hubble tension, the mismatch between measures of how fast the universe expands. Different methods give values that disagree beyond their stated errors. Many groupings of the data have been tried to blame physics or hidden errors.

Their Bayesian jackknife study works through sixteen independent measurements, testing whether they fall into meaningful clusters. After checking the usual early-versus-late split, they try scenarios where certain subsets carry a bias, such as the local distance ladder or microwave background data. Population parameters describe each subset, letting the data itself point to the tension's origin.

Catalog record

CREATORThomas Hughes, Michael J. Wilensky, Philip Bull
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE67 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-19341

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Systematic assessment of the Hubble tension via Bayesian jackknife testing 2026 Systematic assessment of the Hubble tension via Bayesian jackknife testing – A Thomas Hughes, Michael J. Wilensky, Philip Bull Abstract Statistically-significant differences in the value of the Hubble parameter are found depending on the measurement method that is used, a result known as the Hubble tension. A variety of ways of comparing, grouping, and excluding measurements have been used to try to explain this, either in terms of physical effects or systematic errors. We present a systematic ‘Bayesian jackknife’ analysis of 16 independent measurements of the Hubble parameter in an attempt to identify whether the measurements fall into meaningful clusters that would help explain the origin of the tension. After evaluating evidence ratios for the commonly-used split into early- vs late-time measurements, we then study a range of simplified alternative physical scenarios that reflect different physical origins of an apparent bias or shift in the value of H 0 H_{0} , assigning phenomenological population parameters to each subset. These include scenarios where specific subsets are biased (e.g. due to unrecognised experimental systematics in the local distance ladder or cosmic microwave background measurements), as well as more cosmologically-motivated cases involving modifications to the expansion history. Many of these scenarios have similar marginal likelihood, but the model where no measurements are biased is strongly disfavoured. Finally, we marginalise over all these scenarios to estimate the ‘model agnostic’ posterior distribution of H 0 H_{0} . The resulting distribution is mildly multi-modal, but modestly favours values near H 0 = 68 H_{0}=68 km s − 1 Mpc − 1 \mathrm{km}\text{\,}{\mathrm{s}}^{-1}\text{\,}{\mathrm{Mpc}}^{-1} , with a 95% credible region of 66.7 < H 0 < 72.7 66.7<H_{0}<72.7 km s − 1 Mpc − 1 \mathrm{km}\text{\,}{\mathrm{s}}^{-1}\text{\,}{\mathrm{Mpc}}^{-1} . 1 Introduction Recent observational advances have led to increasingly precise cosmological parameter measurements. Rather than resolving open questions, such as the nature of dark matter or dark energy, this progress has resulted in discrepancies and inconsistencies however, most notably a significant tension between different measurements of the Hubble parameter H_{0} . Local (late Universe) determinations using Cepheid-calibrated supernovae yield values of H_{0}=73.04\pm 1.04 \mathrm{km}\text{\,}{\mathrm{s}}^{-1}\text{\,}{\mathrm{Mpc}}^{-1} ( 43 ) or similar, while early Universe estimates inferred from cosmic microwave background (CMB) anisotropies under the \Lambda CDM model are consistent with values closer to H_{0}=67.27\pm 0.6 \mathrm{km}\text{\,}{\mathrm{s}}^{-1}\text{\,}{\mathrm{Mpc}}^{-1} ( 38 ) . The \Lambda CDM model, which assumes a flat universe dominated by cold dark matter and a cosmological constant, has long served as the standard model for interpreting astrophysical data, making this disagreement particularly consequential. This discrepancy, dubbed the “Hubble tension,” has now surpassed the 5\sigma threshold, making it one of the most statistically significant challenges to the standard cosmological model. It remains unresolved despite substantial advances in measurement techniques and calibration precision ( 1 ; 15 ; 27 ; 9 ) . The tension raises a crucial question: is the mismatch indicative of new physics, or does it reflect underestimated systematic errors in some of the datasets? Some have argued that for local measurements, unaccounted for systematics such as peculiar velocities ( 36 ; 7 ) , calibration errors ( 44 ) , or heliocentric redshift uncertainties ( 41 ) , could be inflating estimates of H_{0} . Intrinsic scatter in Type Ia supernova luminosities further complicate the picture; simplified models that attribute this scatter to a single hidden variable may neglect subtler dependencies on dust extinction and host galaxy properties ( 50 ; 35 ; 28 ) , thereby introducing hidden biases in inferred distances. An alternative perspective focuses on issues within the local distance ladder (LDL) itself: while the third rung (Type Ia supernovae) has been highlighted as potentially problematic ( 21 ) , lingering systematics in the earlier calibration steps, parallax and Cepheid distance, may also contribute ( 32 ; 44 ) . Others have suggested that assumptions embedded in early Universe physics, including recombination modelling ( 24 ; 33 ; 4 ) , incomplete sky coverage ( 34 ) , or exotic components such as primordial magnetic fields ( 22 ) , might bias CMB-based inferences. Several theoretical models have also emerged to resolve the tension through new physics. Early Dark Energy (EDE) models, which propose an additional energy component that is dynamically important before recombination, have emerged as one of the most promising avenues, as they allow higher inferred values of H_{0} from early Universe probes without significantly disturbing other cosmological parameters ( 39 ; 10 ; 33 ) . Additional theories involve interactions between dark matter and dark energy, modifications to neutrino properties such as mass hierarchies or sterile species, or the inclusion of novel forms of exotic energy ( 31 ; 24 ; 52 ; 2 ; 25 ; 6 ) . Still, even among these, many models provide only partial relief from the tension, failing to provide a complete resolution. Framing the problem strictly as “early versus late” may be an oversimplification. Several methods, including time-delay lensing ( 30 ) , baryon acoustic oscillations (BAOs) ( 1 ) , and gravitational wave sirens ( 45 ) , depend indirectly on the sound horizon and thus do not fit neatly into either category. As a result, alternative framings have been proposed, such as comparisons based on methodology rather than redshift alone. In this context, a useful distinction emerges within local measurements, which can be split into two categories: LDL approaches and one-step methods. LDL techniques estimate H_{0} by calibrating distances using a chain of calibration ‘rungs’, most commonly using geometric parallax to Cepheids and then SNIa ( 42 ; 43 ) , each introducing potential systematics. In contrast, one-step methods aim to bypass intermediate calibrations entirely and infer H_{0} directly from cosmological phenomena. Some examples include strong lensing time delays ( 30 ) , standard sirens ( 16 ) and the Sunyaev-Zel’dovich (SZ) effect ( 47 ) . The complexity of the Hubble tension has led to a number of statistical approaches being proposed. Initial efforts using tension metrics ( 29 ) , while insightful, showed sensitivity to assumptions about model priors and data independence. These insights build on early foundational work by 40 , who advocated for hierarchical Bayesian approaches in cosmology, emphasising its capacity to propagate uncertainties, incorporate prior information and perform robust model comparisons. More recently, the formalisation of these methods in the context of the Hubble tension has been advanced by 13 , who found that early Universe and late Universe measurements yield significantly different H_{0} values. The tension is ostensibly reduced with this approach to 3\sigma , but still remains strong. In light of this, new Bayesian statistical methods have emerged. 49 introduced CHIBORG , 1 1 1 https://github.com/mwilensky768/chiborg a ‘Bayesian jackknife’ method, which uses Bayesian inference and combinatorics to identify subsets of data that may be responsible for introducing a bias into a measurement. An analysis of 21 cm power spectrum measurements from the Hydrogen Epoch of Reionization Array (HERA) showed how Bayesian tests can improve reliability by validating which data subsets are most consistent with one another ( 20 ) . CHIBORG uses the Bayesian evidence ( 23 , marginal likelihood;) to compare various hypotheses regarding potential biases in the data. In this paper, we extend the work

Read the complete text