Pith. sign in

REVIEW 2 major objections 5 minor 43 references

Climate sensitivity analysis -- A case study from forty years of US compositional cause of death data

T0 review · 2 major / 5 minor · reviewed 2026-07-11 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read US climate extremes raise the share of hypertensive deaths, especially at older ages, with offsets that can hedge life-insurance books.

desk verdict Solid applied CODA-GAM pipeline for US climate-mortality shares; age-specific claims rest on a soft reassignment step, but the between-cause patterns and insurance hedge reading are usable. read the letter →

arxiv 2607.04198 v1 pith:JXCWDZRM submitted 2026-07-05 stat.AP

Michelle Dong , Aaron Bruhn , Francis Hui , Han Lin Shang This is my paper · ORCID
classification stat.AP
keywords compositionaldataanalysiscauseofdeathclimatescenariomodellingclimate-relatedrisksmortalityandforecastinggeneralisedadditivemodelsprincipalcomponent
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks how temperature extremes, sea level, rainfall and related climate factors change the composition of US deaths by cause and age. Death densities are treated as compositional data that must sum to one, so an avoided death from one cause raises the share of another. Climate and macro variables are first reduced by principal components, then linked to the transformed death densities through generalised additive models that allow non-linear effects. Scenario runs that raise high-temperature extremes and sea level while cutting cold extremes show clear rises in the proportion of hypertensive heart-disease deaths, smaller or offsetting moves in other cardiovascular and respiratory causes, and the strongest shifts among ages 55–95. For life insurers those age-and-cause offsets line up with a natural hedge between protection and annuity liabilities under rising climate risk.

What carries the argument

α-transformed compositional data analysis coupled with principal-component reduction of climate/macro covariates and generalised additive models (CODA–GAM): the α-transform maps zero-containing death densities into unconstrained space, the PCs remove collinearity, and the GAM supplies flexible non-linear links that are inverted back to the simplex for scenario interpretation.

What would settle it

Re-run the same CODA–GAM on an independent national cause-of-death series (or on a hold-out decade of US data) after fixing the α-transform and PC loadings; if the reassigned hypertensive share no longer rises systematically with PC1 for ages 55–95, or if the correlation-based label matching assigns the same component to multiple contradictory cause–age cells, the claimed age-and-cause pattern fails.

Watch

Extended reading notes

Core claim

When climate factors are projected onto their first principal component (warmer extremes, higher sea level, fewer cold extremes) and the resulting scores are fed into a CODA–GAM, the share of deaths from hypertensive heart disease rises most sharply, while other climate-linked causes show partial offsets; the largest proportional shifts concentrate in the 55–95 age range and are therefore consistent with a natural hedge between annuity and protection products.

Load-bearing premise

After the inverse transform, each fitted series is reassigned to an original cause–age label solely by highest correlation with the observed trajectory; when neighbouring age bands move almost in lockstep the matching is non-unique and can collapse distinct model components onto the same label.

Editorial extensions

If this is right

  • Stress tests that raise heat and sea-level extremes should produce larger relative increases in hypertensive mortality shares than in ischaemic or COPD shares for ages 55–95.
  • Life-insurance books with both protection and annuity exposure can expect partial natural offsets under climate-driven mortality shifts of the kind modelled here.
  • Scenario design can be refined by stressing PC2 (dry/wet extremes) and PC3 (macro/wind) in addition to PC1, producing joint climate–macro counterfactuals.
  • Age-band differentials imply that climate-related mortality risk is concentrated among older, more vulnerable subgroups rather than uniformly across the population.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the correlation reassignment step is replaced by a supervised matching that preserves cause–age identity, the age-gradient claims could be sharpened or overturned without changing the CODA–GAM itself.
  • The same pipeline applied to multi-country HCD series would test whether the hypertensive offset pattern is US-specific or a broader feature of temperate populations.
  • Because the model works with death proportions rather than absolute rates, combining it with population-exposure projections would convert the compositional scenarios into absolute excess-death numbers usable by regulators.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper models US cause-of-death densities (1979–2023, HCD Intermediate causes, ages 25–95) as compositional data via the α-transformation (α fixed at 0.5), reduces collinear climate (ACI subfactors) and macro covariates by PCA, and fits multivariate GAMs (and linear models) in the transformed space. Sensitivity/scenario analysis varies the first principal component (temperature extremes + sea level) over an extrapolated range and back-transforms to the simplex. The central claims are that climate factors produce cause- and age-specific shifts (notably rising hypertensive-heart proportions with higher PC1), that impacts concentrate at ages 55–95, and that between-cause offsets are consistent with a natural hedge between annuity and protection products for life insurers.

Significance. If the age- and cause-specific patterns hold, the work supplies a transparent, reproducible CODA-based template for climate scenario analysis that life insurers and regulators can use to stress mortality portfolios. Strengths include public code and data, high in-sample Aitchison R² (~0.99), explicit handling of zeros via the α-transformation, and clear partial-effect and scenario plots that make the non-linear PC1 relationships inspectable. The natural-hedge interpretation is practically relevant even if only the coarser between-cause offsets survive scrutiny.

major comments (2)
  1. [Section 4.2] Section 4.2: After the inverse α-transformation, fitted series are reassigned to original cause–age labels solely by highest correlation with the observed trajectories. The paper itself notes that neighbouring age bands within the same cause are highly collinear, so multiple model components can map to the same label and some series (e.g., female I033 age 55) are not uniquely recovered. This reassignment is load-bearing for every claim that effects vary by age within cause, that impacts are more pronounced for ages 55–95, and that within-cause offsetting is weaker than between-cause offsetting. Without a uniqueness check, sensitivity analysis of the matching rule, or an alternative identification strategy (e.g., constrained reassignment or modelling in the original labels), the fine-grained age-gradient statements are not secure.
  2. [Section 3.2] Section 3.2 and the scenario design in Section 4.3: The residual covariance Σ after the α-transformation is assumed diagonal (component-wise independence). With UC−1 ≈ 89 components and only T = 45 years, residual dependence is likely; if present it would affect both the partial-effect confidence bands and the reliability of the extrapolated PC1 scenario curves. The paper flags this as future work, but the assumption is currently load-bearing for the reported uncertainty and for the quantitative scenario statements.
minor comments (5)
  1. [Section 3.1] α is fixed at 0.5 with only a brief statement that other values produce ‘relatively small’ changes. A short sensitivity table (or cross-validated α) would strengthen reproducibility.
  2. [Section 4.3] PC1 is extrapolated to ±8 (roughly double the historical range). The manuscript should state more clearly that the scenario curves are pure extrapolations of the estimated smooths and that no out-of-sample validation is provided.
  3. [Figure 2] Figure 2 caption notes only nine plots for females because of the reassignment collision; this should be flagged earlier in the main text so readers understand why the top-ten set is incomplete.
  4. [Section 2.2 / Appendix A] Table 2 loadings and the correlation matrix (Appendix A) are useful; a brief note on whether the PCA was performed on the full sample or with any hold-out would aid replication.
  5. Minor typographical issues: ‘butad-hocguideline’, ‘inter pretation’, and inconsistent capitalisation of cause names appear in the text and tables.

Circularity Check

1 steps flagged · score 1.5 of 10

Empirical CODA-GAM regression of death shares on climate PCs; scenario curves are genuine extrapolations of estimated smooths, not tautologies. Only soft spot is post-hoc correlation reassignment of inverse-alpha components to cause-age labels.

  1. other [Section 4.2 (Model results), paragraphs describing inverse alpha-transform reassignment]
    "after applying the inverse α-transformation, each predicted series is then reassigned to a cause-age combination by comparing it with the original transformed series. Specifically, for each inverse-transformed prediction, we compute the correlation between its fitted trajectory over time and the trajectories of all original cause–age combinations, and assign the label corresponding to the highest correlation … When this occurs, their transformed series are highly correlated, and so different components of the fitted CODA-GAM model may be matched to the same cause–age combination under the corr"

    The age-specific partial-effect plots and scenario curves that underwrite the claim 'impacts more pronounced for ages 55-95' and 'less offsetting by age within causes' rest on this post-hoc matching. Because the matching is defined by correlation with the very trajectories being labelled, neighbouring collinear age bands can be reassigned non-uniquely; the paper itself records that female I033 age 55 is lost. This is not algebraic circularity of the GAM fit itself, but it makes the fine-grained within-cause age gradients partly self-referential to the observed series rather than independently recovered from the model components.

full rationale

The paper's derivation chain is: (1) form compositional death densities d_t that sum to 1; (2) apply fixed alpha=0.5 transformation to unconstrained space; (3) reduce collinear climate/macro covariates via PCA to three PCs; (4) fit multivariate GAM (or LM) of the transformed responses on the PCs; (5) invert the transformation and re-label components by highest correlation with observed trajectories; (6) vary PC1 over an extrapolated range and read off changes in back-transformed proportions. None of these steps is algebraically forced by a fitted constant or by a self-citation uniqueness theorem. The high Aitchison R2 and public code confirm that the model is an ordinary empirical fit whose predictions are not tautological. The sole circularity-adjacent step is the correlation-based reassignment after the inverse alpha-transform (Section 4.2): when neighbouring age bands within a cause are highly collinear, multiple model components can map to the same label and some series (e.g., female I033 age 55) are not uniquely recovered. This weakens the fine-grained age-gradient claims but does not make the overall compositional response or the direction of PC1 effects on broad causes circular. Fixed alpha=0.5, diagonal Sigma, and PC1 extrapolation are modelling choices already acknowledged by the authors and do not reduce the central claim to its inputs by construction. Score 1.5 reflects one minor interpretive soft spot that is not load-bearing for the between-cause results.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The analysis rests on standard compositional geometry, a fixed Box-Cox-style alpha, three retained principal components, diagonal residual covariance, and a correlation-based label-matching step after back-transformation; no new physical entities are postulated.

free parameters (3)
  • alpha (Box-Cox power in alpha-transformation) = 0.5
    Fixed at 0.5 for all analyses; other values were tested but not optimised by likelihood or cross-validation for the final results.
  • number of principal components retained = 3
    First three PCs kept because they explain >80 % of variance; the 80 % threshold is an ad-hoc convention.
  • PC1 scenario range = [-8, 8]
    Counterfactual interval [-8, +8] chosen as roughly double the historical range; no formal uncertainty quantification accompanies the choice.
assumptions (3)
  • domain assumption Death densities by cause and age form a composition that lives on the simplex and must be transformed before unconstrained regression.
    Stated in Section 3.1 and used throughout; standard CODA premise.
  • ad hoc to paper Residual covariance matrix Sigma after alpha-transformation is diagonal (component-wise independence).
    Explicitly assumed in Section 3.2; authors note it is strong and flag future work to relax it.
  • domain assumption Thin-plate regression splines with default mgcv penalties adequately capture the smooth effects of the three PCs.
    Default settings of Wood (2017) used; alternative bases tested but not reported in detail.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Climate sensitivity analysis -- A case study from forty years of US compositional cause of death data." pith.science (2026). https://pith.science/paper/JXCWDZRM

@misc{pith2026260704198,
  author       = {Pith},
  title        = {Pith review of: Climate sensitivity analysis -- A case study from forty years of US compositional cause of death data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JXCWDZRM}},
  note         = {Machine review of arXiv:2607.04198}
}
read the original abstract

Emerging climate risks pose pressing challenges for insurers, governments, and businesses worldwide, as they face growing uncertainty in quantifying the impacts of climate change from physical risks. This paper aims to understand the impact of specific climate factors on mortality by cause for subgroups of the United States (US) population, through sensitivity and scenario analysis based on an increase in temperature and sea level extremes. We apply compositional data analysis (CODA) techniques to examine cause-specific deaths, treating the density of deaths as a set of dependent, non-negative values that sum to one. We couple CODA with principal component analysis on the climate factors as a means of dimension reduction, and fit generalised additive models to better reflect the non-linear relationships between the dimension-reduced principal scores and mortality by cause. The results of our analysis indicate climate-related factors have varying impacts by cause and ages within each cause, with more pronounced increases in the proportions of deaths from hypertensive heart disease as temperature and sea level extremes increase. Scenario analysis also indicates that an increase in temperature high extremes, sea level, and rainfall, in conjunction with a decrease in low temperature extremes, lead to offsetting impacts on the proportion of deaths between climate-related causes, but less offsetting by age within causes. Furthermore, the impacts on the proportions of death are more pronounced for ages between 55 and 95, reinforcing the observation that climate-related risks have a greater impact on older (and potentially more vulnerable) subgroups of the population. For life insurers specifically, these results are consistent with the natural hedge that arises between annuity and protection products, in light of increasing climate risk.

Figures

Figures reproduced from arXiv: 2607.04198 by the authors.

Figure 1
Figure 1. Proportion of deaths by cause for the US population from 1979 to 2023. The charts on the left (panels a [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Partial effects plots of PC1 based on fitting a CODA-GAM to US females, and depicting the top cause [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Partial effects plots of PC1 based on fitting a CODA-GAM to US males, and depicting the top ten cause [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Scenario analysis for US females, using the fitted CODA-GAM and based on varying PC1 from -8.0 to [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]
Figure 5
Figure 5. Figure 5: Scenario analysis for US males, using the fitted CODA-GAM and based on varying PC1 from -8.0 to 8.0. [PITH_FULL_IMAGE:figures/full_fig_p023_5.png]
Figure 6
Figure 6. Figure 6: Sensitivity analysis for US females, using the CODA-GAM fit, for the top ten combinations of cause and [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]
Figure 7
Figure 7. Figure 7: Sensitivity analysis for US females, using the CODA-GAM fit, for the top ten combinations of cause and [PITH_FULL_IMAGE:figures/full_fig_p025_7.png]
Figure 8
Figure 8. Figure 8: Correlation matrix of explanatory variables for the US data. The size and colour of the circle represent the [PITH_FULL_IMAGE:figures/full_fig_p029_8.png]
Figure 9
Figure 9. Figure 9: GAM R-squared and percentage deviation explained in the [PITH_FULL_IMAGE:figures/full_fig_p031_9.png]
Figure 10
Figure 10. Figure 10: GAM fit for US females and males, top ten causes and age bands. [PITH_FULL_IMAGE:figures/full_fig_p032_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 1 linked inside Pith

  1. [1]

    Actuaries climate index

    Actuaries Climate Index (2025). Actuaries climate index. https://actuariesclimateindex.org. https://actuariesclimateindex.org

    2025

  2. [2]

    Aitchison, J. (1986). The Statistical Analysis of Compositional Data . Monographs on Statistics and Applied Probability. Chapman & Hall, London. Reprinted in 2003 with additional material by The Blackburn Press

    1986

  3. [3]

    Climate scenario considerations for u.s

    American Academy of Actuaries (2025). Climate scenario considerations for u.s. own risk solvency assessment

    2025

  4. [4]

    Australia’s national climate risk assessment report

    Australian Climate Service (2025). Australia’s national climate risk assessment report. 280pp. Available at: https://www.acs.gov.au/pages/national-climate-risk-assessment

    2025

  5. [5]

    Key elements of the 2021 biennial exploratory scenario: Financial risks from climate change

    Bank of England (2021). Key elements of the 2021 biennial exploratory scenario: Financial risks from climate change

    2021

  6. [6]

    Bergeron-Boucher, M.-P., Canudas-Romo, V., Oeppen, J., and Vaupel, J. W. (2017). Coherent forecasts of mortality with compositional data analysis. Demographic Research , 37:527--566

    2017

  7. [7]

    Bergeron-Boucher, M.-P., Strozza, C., Simonacci, V., and Oeppen, J. (2025). Modeling and forecasting healthy life expectancy with compositional data analysis. Demography , 62(3):787--810

    2025

  8. [8]

    A., Boudreault, J., and Raymond, S

    Bourget, M., Boudreault, M., Carozza, D. A., Boudreault, J., and Raymond, S. (2024). A data science approach to climate change risk assessment applied to pluvial flood occurrences for the U nited S tates and C anada. ASTIN Bulletin: The Journal of the IAA , 54(3):495--517

    2024

Show all 43 references
  1. [9]

    Box, G. E. and Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical Society: Series B (Methodological) , 26(2):211--243

    1964

  2. [10]

    M., Shang, H

    Dong, Z. M., Shang, H. L., Hui, F., and Bruhn, A. (2025). A compositional approach to modeling cause-specific mortality with zero counts. Annals of Actuarial Science , 19(3):416--441

    2025

  3. [11]

    Gao, Y., Zhang, Z., Cai, Z., Zhu, X., Zou, T., and Wang, H. (2025). Penalized sparse covariance regression with high dimensional covariates . Journal of Business & Economic Statistics , 43:615--626

    2025

  4. [12]

    Gasparrini, A., Guo, Y., Hashizume, M., Lavigne, E., Zanobetti, A., Schwartz, J., Tobias, A., Tong, S., Rockl \"o v, J., Forsberg, B., et al. (2015). Mortality risk attributable to high and low ambient temperature: A multicountry observational study. The Lancet , 386(9991):369--375

    2015

  5. [13]

    M., Huber, V., Tong, S., Coelho, M

    Gasparrini, A., Guo, Y., Sera, F., Vicedo-Cabrera, A. M., Huber, V., Tong, S., Coelho, M. d. S. Z. S., Saldiva, P. H. N., Lavigne, E., Correa, P. M., et al. (2017). Projections of temperature-related excess mortality under climate change scenarios. The Lancet Planetary Health ...

    2017

  6. [14]

    Greenacre, M. (2021). Compositional data analysis. Annual Review of Statistics and its Application , 8:271--299

    2021

  7. [15]

    J., Hastie, T., d’Enza, A

    Greenacre, M., Groenen, P. J., Hastie, T., d’Enza, A. I., Markos, A., and Tuzhilina, E. (2022). Principal component analysis. Nature Reviews Methods Primers , 2(1):100

    2022

  8. [16]

    Harezlak, J., Ruppert, D., and Wand, M. P. (2018). Semiparametric regression with R , volume 109. Springer

    2018

  9. [17]

    and Satten, G

    Hu, Y.-J. and Satten, G. A. (2023). Compositional analysis of microbiome data using the linear decomposition model (ldm). Bioinformatics , 39(11):btad668

    2023

  10. [18]

    G., Wang, X., Vaneckova, P., FitzGerald, G., and Tong, S

    Huang, C., Barnett, A. G., Wang, X., Vaneckova, P., FitzGerald, G., and Tong, S. (2011). Projecting future heat-related mortality under climate change scenarios: A systematic review. Environmental health perspectives , 119(12):1681--1690

    2011

  11. [19]

    Hui, F. K. C., Maestrini, L., and Welsh, A. H. (2024). Homogeneity pursuit and variable selection in regression models for multivariate abundance data . Biometrics , 80:ujad001

    2024

  12. [20]

    French Institute for Demographic Studies (France), Max Planck Institute for Demographic Research (Germany) and the University of California, Berkeley (USA)

    Human Cause-of-Death Data Series (2025). French Institute for Demographic Studies (France), Max Planck Institute for Demographic Research (Germany) and the University of California, Berkeley (USA). A vailable at: https://www.mortality.org/Data/HCD (accessed June 25, 2024)

    2025

  13. [21]

    Importance of climate-related risks for actuaries

    IAA: Climate Risk Task Force (2020). Importance of climate-related risks for actuaries. URL: https://www.actuaries.org/IAA/Documents/Publications/Papers/CRTF_Paper4_Asset_Portolios.pdf

    2020

  14. [22]

    Introduction to climate-related scenarios

    IAA: Climate Risk Task Force (2021). Introduction to climate-related scenarios. https://actuaries.org/app/uploads/2025/04/CRTF_Introduction_Climate_Scenarios.pdf

    2021

  15. [23]

    Application of climate-related risk scenarios to asset portfolios

    IAA: Climate Risk Task Force (2022). Application of climate-related risk scenarios to asset portfolios. URL: https://www.actuaries.org/IAA/Documents/Publications/Papers/CRTF_Paper4_Asset_Portolios.pdf

    2022

  16. [24]

    The emperor's new climate scenarios

    Institute and Faculty of Actuaries (2023). The emperor's new climate scenarios

    2023

  17. [25]

    Jolliffe, I. T. and Cadima, J. (2016). Principal component analysis: A review and recent developments. Philosophical transactions of the royal society A: Mathematical, Physical and Engineering Sciences , 374(2065):20150202

    2016

  18. [26]

    E., Bergeron-Boucher, M.-P., Oeppen, J., and Kallestrup-Lamb, M

    Kjaergaard, S., Ergemen, Y. E., Bergeron-Boucher, M.-P., Oeppen, J., and Kallestrup-Lamb, M. (2020). Longevity forecasting by socio-economic groups using compositional data analysis. Journal of the Royal Statistical Society: Series A (Statistics in Society) , 183(3):1167--1187

    2020

  19. [27]

    E., Kallestrup-Lamb , M., Oeppen, J., and Lindahl-Jacobsen , R

    Kjaergaard, S., Ergemen, Y. E., Kallestrup-Lamb , M., Oeppen, J., and Lindahl-Jacobsen , R. (2019). Forecasting causes of death using compositional data analysis: the case of cancer deaths. Journal of the Royal Statistical Society Series C: Applied Statistics , 68(5):1351--1370

    2019

  20. [28]

    M., Prifti, K., Bell, M

    Lee, W., Kim, Y., Sera, F., Gasparrini, A., Park, R., Choi, H. M., Prifti, K., Bell, M. L., Abrutzky, R., Guo, Y., et al. (2020). Projections of excess mortality related to diurnal temperature range under climate change scenarios: A multi-country modelling study. The Lancet Pl...

    2020

  21. [29]

    CP10/25 – Enhancing banks’ and insurers’ approaches to managing climate-related risks – Update to SS3/19

    Prudential Regulation Authority (2025a). CP10/25 – Enhancing banks’ and insurers’ approaches to managing climate-related risks – Update to SS3/19 . https://www.bankofengland.co.uk/prudential-regulation/publication/2025/april/enhancing-banks-and-insurers-approaches-to-managing-...

    2025

  22. [30]

    PS25/25 – Enhancing banks’ and insurers’ approaches to managing climate-related risks – Update to SS3/19

    Prudential Regulation Authority (2025b). PS25/25 – Enhancing banks’ and insurers’ approaches to managing climate-related risks – Update to SS3/19 . https://www.bankofengland.co.uk/prudential-regulation/publication/2025/december/enhancing-banks-and-insurers-approaches-to-managi...

    2025

  23. [31]

    The risk of a lifetime: mapping the impact of climate change on life and health risks

    Swiss Re Institute (2023). The risk of a lifetime: mapping the impact of climate change on life and health risks

    2023

  24. [32]

    PRA Climate Change Adaptation Report 2025

    The Bank of England (2025). PRA Climate Change Adaptation Report 2025 . https://www.bankofengland.co.uk/prudential-regulation/publication/2025/january/pra-climate-change-adaptation-report-2025

    2025

  25. [33]

    and Athineou, G

    Tsagris, M. and Athineou, G. (2025). Compositional: Compositional Data Analysis . R package version 8.0. Available at: https://CRAN.R-project.org/package=Compositional

    2025

  26. [34]

    and Stewart, C

    Tsagris, M. and Stewart, C. (2022). A review of flexible transformations for modeling compositional data. In He, W., Wang, L., Chen, J., and Lin, C. D., editors, Advances and Innovations in Statistics and Data Science , pages 225--234. Springer

    2022

  27. [35]

    T., Preston, S., and Wood, A

    Tsagris, M. T., Preston, S., and Wood, A. T. (2011). A data-based power transformation for compositional data. In Proceedings of the 4th international workshop on Compositional Data Analysis , Girona, Spain. URL: https://arxiv.org/abs/1106.1451

    2011 arXiv

  28. [36]

    Climate change indicators: heat-related deaths

    United States Environmental Protection Agency (2025). Climate change indicators: heat-related deaths . https://www.epa.gov/climate-indicators/climate-change-indicators-heat-related-deaths

    2025

  29. [37]

    The coming deluge: Scenario analysis for underwriting in a changing climate

    University of Cambridge Institute for Sustainability Leadership (2022). The coming deluge: Scenario analysis for underwriting in a changing climate

    2022

  30. [38]

    Wood, S. N. (2017). Generalized additive models: an introduction with R . Chapman and Hall/CRC

    2017

  31. [39]

    United states data

    World Bank (2025). United states data. https://data.worldbank.org/country/united-states. Accessed June 13, 2025

    2025

  32. [40]

    Climate change fact sheet

    World Health Organisation (2023). Climate change fact sheet . Accessed 11 June 2025: https://www.who.int/news-room/fact-sheets/detail/climate-change-and-health

    2023

  33. [41]

    The ICD-10 classification of mental and behavioural disorders

    World Health Organization (1992). The ICD-10 classification of mental and behavioural disorders. URL: https://cdn.who.int/media/docs/default-source/classification/other-classifications/9241544228_eng.pdf

    1992

  34. [42]

    World Health Organization et al. (2014). Quantitative risk assessment of the effects of climate change on selected causes of death, 2030s and 2050s. https://iris.who.int/server/api/core/bitstreams/785da4b3-7797-44f5-a01b-eab29c3ac99b/content

    2014

  35. [43]

    M., et al

    Zhao, Q., Guo, Y., Ye, T., Gasparrini, A., Tong, S., Overcenco, A., Urban, A., Schneider, A., Entezari, A., Vicedo-Cabrera, A. M., et al. (2021). Global, regional, and national burden of mortality associated with non-optimal ambient temperatures from 2000 to 2019: A three-stag...

    2021

Pith tools

Reviewed July 11, 2026 · model on record in the stance chip above.