pith. sign in

arxiv: 2606.11715 · v1 · pith:LWIPN2H2new · submitted 2026-06-10 · 📊 stat.ME

Bracketing Relationships of Weighted Average Treatment Effects

Pengfei Tian , Fan Yang , Peng Ding This is my paper

Pith reviewed 2026-06-27 09:03 UTC · model grok-4.3

classification 📊 stat.ME
keywords causal inferenceoverlap weightsaverage treatment effectpropensity scorebracketingobservational studiesinstrumental variablesconditional average treatment effect
0
0 comments X

The pith

The overlap-weighted average treatment effect lies between the effects on the treated and on the control when propensity score and conditional average treatment effect are monotonically related.

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

The paper establishes that, under a monotonic link between the propensity score and the conditional average treatment effect, the average treatment effect computed with overlap weights falls between the average treatment effect on the treated and the average treatment effect on the control. It extends the same bracketing result to weighted local average treatment effects with a binary instrument and to several other weighting schemes. The finding supplies a theoretical ordering among common weighting estimators used in observational causal studies. A practical recommendation follows: plot the estimated conditional average treatment effect against the estimated propensity score to inspect the monotonicity condition directly.

Core claim

Under the canonical setting of observational studies, the average treatment effect under the overlap weight is bounded between the average treatment effects on the treated and control when the propensity score and the conditional average treatment effect satisfy a monotonic relationship. The same bracketing holds for weighted local average treatment effects in the binary-instrument binary-treatment setting and extends to other common weights.

What carries the argument

The bracketing inequality that places the overlap-weighted ATE between ATT and ATC under monotonicity of propensity score with conditional average treatment effect.

If this is right

  • Overlap weights produce an estimate that is theoretically intermediate between the two subgroup effects.
  • The bracketing property carries over to other weighting functions once the monotonicity condition is imposed.
  • The CP-plot of conditional average treatment effect against propensity score serves as a direct diagnostic for the assumption.
  • The ordering supplies a consistency check across different weighted estimators in the same study.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The result may encourage routine reporting of all three quantities (overlap, treated, control) together rather than choosing one weight in isolation.
  • If monotonicity is implausible, the ordering itself can be used as a sensitivity probe for hidden confounding.
  • The same logic could be checked in simulation studies that deliberately violate monotonicity to map the size of possible reversals.

Load-bearing premise

The propensity score and the conditional average treatment effect move monotonically together.

What would settle it

A data set in which the estimated overlap-weighted ATE falls outside both the ATT and ATC estimates even though the scatter of estimated conditional average treatment effect versus estimated propensity score shows no monotonic pattern.

Figures

Figures reproduced from arXiv: 2606.11715 by Fan Yang, Peng Ding, Pengfei Tian.

Figure 1
Figure 1. Figure 1: CP-plots: estimated conditional treatment effects ˆτ [PITH_FULL_IMAGE:figures/full_fig_p018_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Estimated CATE among compliers ˆτ c (X) versus estimated IV propensity scores eˆ(X) in the 401(k) application. Points are unit-level modeled conditional Wald estimates; the solid curve is a weighted LOESS smoother with a pointwise confidence band. We next estimate the corresponding local weighted IV estimands. For a generic weight h(X), we focus on the local weighted causal estimand defined in Section 3, τ… view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of Theorem 4. The notation “→” stands for “≤” under Assump￾tion 1(i) and for “≥” under Assumption 1(ii). Under the linear specification in Assumption 2, adjacent beta-weighted estimands satisfy a convex-combination relation. Proposition 3 Assume e(X) ∈ (0, 1) and Assumption 2. Then τu+1,v+1 = λ · τu+1,v + (1 − λ) · τu,v+1, (6) where λ ∈ [0, 1]. When e(X) is constant, we have τu+1,v+1 = τu+1,v… view at source ↗
read the original abstract

Under the canonical setting of observational studies for causal inference, we show that the average treatment effect under the overlap weight, the weight that is proportional to the conditional variance of the treatment given the covariates, is bounded between the average treatment effects on the treated and control, under a monotonic relationship between the propensity score and the conditional average treatment effect. We further extend the result to weighted local average treatment effects, under the canonical setting with a binary instrumental variable and a binary treatment. We also extend the results to other weights. Based on the theory, we recommend the ``CP-plot'' of the estimated conditional average treatment effect against the estimated propensity score.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript establishes that, under a monotonicity assumption between the propensity score p(x) and the conditional average treatment effect au(x), the overlap-weighted ATE (with weights proportional to p(x)(1-p(x))) is bounded between the ATT and ATC. It derives this via expressing each quantity as a weighted average of au(x) and applying the rearrangement inequality (or equivalent covariance argument), extends the result to weighted LATEs under binary IV and binary treatment, generalizes to other weights, and recommends the CP-plot of estimated CATE versus estimated propensity score for empirical assessment of monotonicity.

Significance. If the central bracketing result holds, it supplies a theoretically grounded relationship among common weighted ATE estimators that can aid interpretation and sensitivity analysis in observational causal inference. The explicit statement of the monotonicity assumption, the use of the rearrangement inequality for the derivation, and the practical CP-plot recommendation are strengths; the extension to the IV setting broadens applicability.

major comments (2)
  1. [§3, Theorem 1] §3, Theorem 1: the bracketing inequality is derived under the stated monotonicity between p(x) and au(x), but the manuscript does not explicitly verify that the overlap weights w(x) = p(x)(1-p(x)) preserve the ordering needed for the rearrangement inequality when p(x) is estimated rather than known; this step is load-bearing for the claim that the result applies to feasible estimators.
  2. [§4.2, Eq. (12)] §4.2, Eq. (12): the extension to weighted LATE replaces the propensity score with the instrument propensity, yet the monotonicity condition is restated without showing that the binary IV exclusion restriction is compatible with the required ordering of the conditional LATE; a counter-example or additional assumption would strengthen the claim.
minor comments (2)
  1. [Abstract] The abstract states the result but omits any mention of the rearrangement inequality or the explicit form of the monotonicity assumption, making the central contribution harder to assess at first reading.
  2. [§5] The CP-plot is introduced in §5 without a formal definition of the plotted quantities or guidance on sample-size requirements for reliable visual assessment of monotonicity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and the recommendation for minor revision. We address the two major comments point by point below.

read point-by-point responses

  1. Referee: [§3, Theorem 1] §3, Theorem 1: the bracketing inequality is derived under the stated monotonicity between p(x) and τ(x), but the manuscript does not explicitly verify that the overlap weights w(x) = p(x)(1-p(x)) preserve the ordering needed for the rearrangement inequality when p(x) is estimated rather than known; this step is load-bearing for the claim that the result applies to feasible estimators.

    Authors: Theorem 1 is a population-level result that applies the rearrangement inequality directly to the true propensity score p(x) and the true overlap weights w(x) = p(x)(1-p(x)) under the stated monotonicity. The manuscript centers on this theoretical relationship among the population quantities. For feasible estimators, the bracketing holds asymptotically under standard consistency conditions on the estimators of the weighted ATEs. We will add a clarifying remark after Theorem 1 to distinguish the population statement from its implications for estimated quantities. revision: yes

  2. Referee: [§4.2, Eq. (12)] §4.2, Eq. (12): the extension to weighted LATE replaces the propensity score with the instrument propensity, yet the monotonicity condition is restated without showing that the binary IV exclusion restriction is compatible with the required ordering of the conditional LATE; a counter-example or additional assumption would strengthen the claim.

    Authors: The exclusion restriction is an identifying assumption required for the IV to isolate the LATE and is logically separate from the monotonicity assumption between the instrument propensity and the conditional LATE. The latter is an additional modeling assumption on effect heterogeneity that can coexist with exclusion in standard binary IV settings. We will insert a brief paragraph in §4.2 noting that the two assumptions operate on distinct aspects of the model and do not conflict under the maintained IV framework. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The derivation expresses ATT, ATC, and overlap-weighted ATE as weighted averages of the CATE tau(x), then applies the rearrangement inequality (or covariance sign argument) under the paper's explicitly stated monotonicity assumption between p(x) and tau(x). This is a standard, externally verifiable mathematical step with no fitted inputs renamed as predictions, no self-definitional loops, and no load-bearing self-citations. The result is independent of the paper's own data or parameters and holds conditionally on the stated assumption, which is assessed separately via the recommended CP-plot.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the canonical observational-study setting plus the monotonicity assumption between propensity score and CATE. No free parameters or new entities are introduced in the abstract.

axioms (1)
  • domain assumption monotonic relationship between the propensity score and the conditional average treatment effect
    This is the explicit condition required for the bounding result to hold, as stated in the abstract.

pith-pipeline@v0.9.1-grok · 5624 in / 1167 out tokens · 23741 ms · 2026-06-27T09:03:08.223790+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

300 extracted references · 6 linked inside Pith

  1. [1]

    2024 , publisher=

    A first course in causal inference , author=. 2024 , publisher=

    2024

  2. [2]

    arXiv preprint arXiv:2601.22572 , year=

    Propensity score weighted Cox regression for survival outcomes in observational studies with multiple or factorial treatments , author=. arXiv preprint arXiv:2601.22572 , year=

    arXiv

  3. [3]

    and Tang, Haihan and Zhang, Yichong , title =

    Jiang, Liang and Linton, Oliver B. and Tang, Haihan and Zhang, Yichong , title =. The Review of Economics and Statistics , volume =. 2026 , month =

    2026

  4. [4]

    Biometrika , volume=

    Instrumental variables as bias amplifiers with general outcome and confounding , author=. Biometrika , volume=. 2017 , publisher=

    2017

  5. [5]

    JAMA , volume=

    The effectiveness of right heart catheterization in the initial care of critically ill patients , author=. JAMA , volume=

  6. [6]

    Biometrics , volume=

    Doubly-robust dynamic treatment regimen estimation via weighted least squares , author=. Biometrics , volume=. 2015 , publisher=

    2015

  7. [7]

    , year =

    Harrell, Frank E., Jr. , year =

  8. [8]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Signed directed acyclic graphs for causal inference , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2010 , publisher=

    2010

  9. [9]

    Biometrics , volume=

    The sign of the bias of unmeasured confounding , author=. Biometrics , volume=. 2008 , publisher=

    2008

  10. [10]

    Review of Economics and Statistics , volume=

    Interpreting OLS estimands when treatment effects are heterogeneous: Smaller groups get larger weights , author=. Review of Economics and Statistics , volume=. 2022 , publisher=

    2022

  11. [11]

    Biometrical Journal , volume=

    Causal inference in the absence of positivity: The role of overlap weights , author=. Biometrical Journal , volume=. 2024 , publisher=

    2024

  12. [12]

    Journal of Causal Inference , volume=

    Bounds on the fixed effects estimand in the presence of heterogeneous assignment propensities , author=. Journal of Causal Inference , volume=. 2025 , publisher=

    2025

  13. [13]

    arXiv preprint arXiv:2411.00947 , year=

    Asymptotic theory for the quadratic assignment procedure , author=. arXiv preprint arXiv:2411.00947 , year=

    arXiv

  14. [14]

    arXiv preprint arXiv:2503.13986 , year=

    Stratified Permutational Berry--Esseen Bounds and Their Applications to Statistics , author=. arXiv preprint arXiv:2503.13986 , year=

    arXiv

  15. [15]

    2011 , publisher=

    Normal approximation by Stein's method , author=. 2011 , publisher=

    2011

  16. [16]

    Bernoulli , pages=

    On the error bound in a combinatorial central limit theorem , author=. Bernoulli , pages=. 2015 , publisher=

    2015

  17. [17]

    Theory of Probability & Its Applications , volume=

    On a combinatorial limit theorem , author=. Theory of Probability & Its Applications , volume=. 1974 , publisher=

    1974

  18. [18]

    Quantitative Economics , year=

    A Robust Permutation Test for Subvector Inference in Linear Regressions , author=. Quantitative Economics , year=

  19. [19]

    Journal of the American Statistical Association , volume=

    Inference in experiments with matched pairs , author=. Journal of the American Statistical Association , volume=. 2022 , publisher=

    2022

  20. [20]

    Optimal Stratification of Survey Experiments , author=

  21. [21]

    arXiv preprint arXiv:2307.15181 , year=

    On the efficiency of finely stratified experiments , author=. arXiv preprint arXiv:2307.15181 , year=

    arXiv

  22. [22]

    arXiv preprint arXiv:2205.02726 , year=

    Asymptotic efficiency bounds for a class of experimental designs , author=. arXiv preprint arXiv:2205.02726 , year=

    Pith/arXiv arXiv

  23. [23]

    Econometrica , pages=

    On the role of the propensity score in efficient semiparametric estimation of average treatment effects , author=. Econometrica , pages=. 1998 , publisher=

    1998

  24. [24]

    Theory of Probability & Its Applications , volume=

    On limit distributions of a statistic , author=. Theory of Probability & Its Applications , volume=. 1975 , publisher=

    1975

  25. [25]

    Ho, Soo-Thong and Chen, Louis HY , journal=. An. 1978 , publisher=

    1978

  26. [26]

    Journal of the American statistical association , volume=

    Estimates of the regression coefficient based on Kendall's tau , author=. Journal of the American statistical association , volume=. 1968 , publisher=

    1968

  27. [27]

    Statistical aspects of

    Panaretos, Victor M and Zemel, Yoav , journal=. Statistical aspects of. 2019 , publisher=

    2019

  28. [28]

    Probability Surveys , volume=

    Fundamentals of Stein’s method , author=. Probability Surveys , volume=

  29. [29]

    Shi, Lei and Ding, Peng , journal=. Berry--

  30. [30]

    2007 , volume=

    Goldstein, Larry , journal=. 2007 , volume=

    2007

  31. [31]

    arXiv preprint arXiv:0912.0726 , year=

    New estimates of the convergence rate in the Lyapunov theorem , author=. arXiv preprint arXiv:0912.0726 , year=

    Pith/arXiv arXiv

  32. [32]

    2010 , publisher=

    Probability theory: an analytic view , author=. 2010 , publisher=

    2010

  33. [33]

    Proceedings of the National Academy of Sciences , volume=

    Improving massive experiments with threshold blocking , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

    2016

  34. [34]

    2015 , journal=

    Rates of convergence for multivariate normal approximation with applications to dense graphs and doubly indexed permutation statistics , author=. 2015 , journal=

    2015

  35. [35]

    arXiv preprint arXiv:1003.6039 , year=

    Stein couplings for normal approximation , author=. arXiv preprint arXiv:1003.6039 , year=

    Pith/arXiv arXiv

  36. [36]

    The Annals of Statistics , volume=

    Rerandomization with diminishing covariate imbalance and diverging number of covariates , author=. The Annals of Statistics , volume=. 2022 , publisher=

    2022

  37. [37]

    The Annals of Statistics , volume=

    Stein’s method of normal approximation: Some recollections and reflections , author=. The Annals of Statistics , volume=. 2021 , publisher=

    2021

  38. [38]

    Zeitschrift f

    An estimate of the remainder in a combinatorial central limit theorem , author=. Zeitschrift f. 1984 , publisher=

    1984

  39. [39]

    Alea , volume=

    Multivariate normal approximation using exchange-able pairs , author=. Alea , volume=

  40. [40]

    Journal of Statistical Planning and Inference , volume=

    Normal approximation to the hypergeometric distribution in nonstandard cases and a sub-Gaussian Berry--Esseen theorem , author=. Journal of Statistical Planning and Inference , volume=. 2007 , publisher=

    2007

  41. [41]

    Journal of the American Statistical Association , volume=

    Post-stratification: a modeler's perspective , author=. Journal of the American Statistical Association , volume=. 1993 , publisher=

    1993

  42. [42]

    Journal of the American Statistical Association , volume=

    Estimation employing post strata , author=. Journal of the American Statistical Association , volume=. 1966 , publisher=

    1966

  43. [43]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Assumption-lean inference for generalised linear model parameters , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2022 , publisher=

    2022

  44. [44]

    Zeitschrift f

    Remainder term estimate in a combinatorial limit theorem , author=. Zeitschrift f. 1976 , publisher=

    1976

  45. [45]

    Journal of the American Statistical Association , volume=

    Balancing covariates via propensity score weighting , author=. Journal of the American Statistical Association , volume=. 2018 , publisher=

    2018

  46. [46]

    American journal of epidemiology , volume=

    Addressing extreme propensity scores via the overlap weights , author=. American journal of epidemiology , volume=. 2019 , publisher=

    2019

  47. [47]

    Proceedings of the sixth Berkeley symposium on mathematical statistics and probability, volume 2: Probability theory , volume=

    A bound for the error in the normal approximation to the distribution of a sum of dependent random variables , author=. Proceedings of the sixth Berkeley symposium on mathematical statistics and probability, volume 2: Probability theory , volume=. 1972 , organization=

    1972

  48. [48]

    Specification and estimation of simultaneous equation models , Volume =

    Hausman, Jerry A , Date-Added =. Specification and estimation of simultaneous equation models , Volume =. Handbook of econometrics , Pages =

  49. [49]

    Statistical Science , Title =

    Imbens, Guido , Date-Added =. Statistical Science , Title =

  50. [50]

    Principal stratification in causal inference , Volume =

    Frangakis, Constantine E and Rubin, Donald B , Date-Added =. Principal stratification in causal inference , Volume =. Biometrics , Pages =

  51. [51]

    The distribution of the ratio of jointly normal variables , Volume =

    Cedilnik, Anton and Kosmelj, Katarina and Blejec, Andrej , Date-Added =. The distribution of the ratio of jointly normal variables , Volume =. Metodolo

  52. [52]

    Three occurrences of the hyperbolic-secant distribution , Volume =

    Ding, Peng , Date-Added =. Three occurrences of the hyperbolic-secant distribution , Volume =. The American Statistician , Number =

  53. [53]

    arXiv preprint arXiv:1606.04146 , Title =

    Kang, Hyunseung and Peck, Laura and Keele, Luke , Date-Added =. arXiv preprint arXiv:1606.04146 , Title =

    Pith/arXiv arXiv

  54. [54]

    A comparison of methods for estimating the causal effect of a treatment in randomized clinical trials subject to noncompliance , Volume =

    Little, Roderick J and Long, Qi and Lin, Xihong , Date-Added =. A comparison of methods for estimating the causal effect of a treatment in randomized clinical trials subject to noncompliance , Volume =. Biometrics , Number =

  55. [55]

    Compliance as an explanatory variable in clinical trials , Volume =

    Efron, Bradley and Feldman, David , Date-Added =. Compliance as an explanatory variable in clinical trials , Volume =. Journal of the American Statistical Association , Number =

  56. [56]

    More powerful randomization-based p-values in double-blind trials with non-compliance , Volume =

    Rubin, Donald B , Date-Added =. More powerful randomization-based p-values in double-blind trials with non-compliance , Volume =. Statistics in medicine , Number =

  57. [57]

    The fitting of straight lines if both variables are subject to error , Volume =

    Wald, Abraham , Date-Added =. The fitting of straight lines if both variables are subject to error , Volume =. The Annals of Mathematical Statistics , Number =

  58. [58]

    2024 , note =

    causaldata: Example Data Sets for Causal Inference Textbooks , author =. 2024 , note =

    2024

  59. [59]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Principal stratification analysis using principal scores , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2017 , publisher=

    2017

  60. [60]

    Instrumental variable methods for causal inference: early work and recent developments , Year =

    Baker, Stuart G , Date-Added =. Instrumental variable methods for causal inference: early work and recent developments , Year =. Statistics in Medicine , Number =

  61. [61]

    Simultaneous-equation estimation in a clinical trial of the effect of smoking on birth weight , Year =

    Permutt, Thomas and Hebel, J Richard , Date-Added =. Simultaneous-equation estimation in a clinical trial of the effect of smoking on birth weight , Year =. Biometrics , Pages =

  62. [62]

    and EDWARDS, ROBERT and SEGNAN, NEREO , Date-Added =

    Cuzick, J. and EDWARDS, ROBERT and SEGNAN, NEREO , Date-Added =. Adjusting for non-compliance and contamination in randomized clinical trials , Volume =. Statistics in Medicine , Number =

  63. [63]

    On estimating efficacy from clinical trials , Volume =

    Sommer, Alfred and Zeger, Scott L , Date-Added =. On estimating efficacy from clinical trials , Volume =. Statistics in medicine , Number =

  64. [64]

    The paired availability design: a proposal for evaluating epidural analgesia during labor , Volume =

    Baker, Stuart G and Lindeman, Karen S , Date-Added =. The paired availability design: a proposal for evaluating epidural analgesia during labor , Volume =. Statistics in medicine , Number =

  65. [65]

    Accounting for no-shows in experimental evaluation designs , Volume =

    Bloom, Howard S , Date-Added =. Accounting for no-shows in experimental evaluation designs , Volume =. Evaluation review , Number =

  66. [66]

    Randomized consent designs for clinical trials: an update , Volume =

    Zelen, Marvin , Date-Added =. Randomized consent designs for clinical trials: an update , Volume =. Statistics in medicine , Number =

  67. [67]

    A new design for randomized clinical trials , Volume =

    Zelen, Marvin , Date-Added =. A new design for randomized clinical trials , Volume =. New England Journal of Medicine , Number =

  68. [68]

    Assessing the effect of an influenza vaccine in an encouragement design , Volume =

    Hirano, Keisuke and Imbens, Guido W and Rubin, Donald B and Zhou, Xiao-Hua , Date-Added =. Assessing the effect of an influenza vaccine in an encouragement design , Volume =. Biostatistics , Number =

  69. [69]

    Effects of self-study for coachable test item types

    Powers, Donald E and Swinton, Spencer S , Date-Added =. Effects of self-study for coachable test item types. , Volume =. Journal of Educational Psychology , Number =

  70. [70]

    Bayesian inference for causal effects in randomized experiments with noncompliance , Year =

    Imbens, Guido W and Rubin, Donald B , Date-Added =. Bayesian inference for causal effects in randomized experiments with noncompliance , Year =. The Annals of Statistics , Pages =

  71. [71]

    Identification and estimation of local average treatment effects , Volume =

    Imbens, Guido W and Angrist, Joshua D , Date-Added =. Identification and estimation of local average treatment effects , Volume =. Econometrica , Number =

  72. [72]

    Identification of causal effects using instrumental variables , Volume =

    Angrist, Joshua D and Imbens, Guido W and Rubin, Donald B , Date-Added =. Identification of causal effects using instrumental variables , Volume =. Journal of the American Statistical Association , Number =

  73. [73]

    Two-stage least squares estimation of average causal effects in models with variable treatment intensity , Volume =

    Angrist, Joshua D and Imbens, Guido W , Date-Added =. Two-stage least squares estimation of average causal effects in models with variable treatment intensity , Volume =. Journal of the American Statistical Association , Number =

  74. [74]

    The environment and disease: association or causation? , Volume =

    Hill, Austin Bradford , Date-Added =. The environment and disease: association or causation? , Volume =. Proceedings of the Royal society of Medicine , Number =

  75. [75]

    Potential updates to Cornfield's 1959 `Principles of Research' , Volume =

    Rubin, Donald B , Date-Added =. Potential updates to Cornfield's 1959 `Principles of Research' , Volume =. Statistics in medicine , Number =

    1959

  76. [76]

    Generalized Cornfield conditions for the risk difference , Volume =

    Ding, Peng and Vanderweele, Tyler J , Date-Added =. Generalized Cornfield conditions for the risk difference , Volume =. Biometrika , Number =

  77. [77]

    Principles of research

    Cornfield, Jerome , Date-Added =. Principles of research. , Volume =. American journal of mental deficiency , Pages =

  78. [78]

    The significance of cell type in relation to the aetiology of lung cancer , Volume =

    Doll, R and Hill, A Bradford and Kreyberg, L , Date-Added =. The significance of cell type in relation to the aetiology of lung cancer , Volume =. British journal of cancer , Number =

  79. [79]

    Smoking and lung cancer: recent evidence and a discussion of some questions , Volume =

    Cornfield, Jerome and Haenszel, William and Hammond, E Cuyler and Lilienfeld, Abraham M and Shimkin, Michael B and Wynder, Ernst L , Date-Added =. Smoking and lung cancer: recent evidence and a discussion of some questions , Volume =. J. Nat. Cancer Inst , Pages =

  80. [80]

    The direction of time , Volume =

    Reichenbach, Hans , Date-Added =. The direction of time , Volume =

Showing first 80 references.