Nested Sensitivity Envelopes for Transported Quantile Treatment Effects
Pengyun Wang · 2026
Read Nested Sensitivity Envelopes for Transported Quantile Treatment Effects by Pengyun Wang (2026) 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-2605-09264.txt.
Introduction
Wang studies quantile treatment effects for a target population when a source study has unmeasured confounding and may not transport after conditioning on observed covariates. Data pair a source sample holding treatment, outcome, and covariates with a target sample holding covariates only. Two sensitivity restrictions apply: an odds-ratio bound and a likelihood-ratio bound on outcome shift.
For each arm and threshold, Wang derives a sharp target counterfactual CDF envelope. It nests a source sensitivity map within a target outcome-shift map. The envelopes are sharp at process level, attainable as entire CDFs, and invert to sharp quantile and QTE bounds. Semiparametric theory follows, with a canonical gradient, cross-fitted one-step estimators, and Gaussian approximation on regular index sets. Full index sets use Hadamard differentiability and subsampling-valid inference. Inverted monotone CDF bands yield honest confidence sets, and a breakdown frontier frames non-refutation as level sets.
Catalog record
| CREATOR | Pengyun Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 133 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.09264
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Nested Sensitivity Envelopes for Transported Quantile Treatment Effects Pengyun Wang Abstract We study target-population quantile treatment effects when a source study may have unmeasured treatment confounding and may not transport to a target population after conditioning on observed covariates. The observed data consist of a source sample with treatment, outcome and covariates, and a target sample with covariates only. We impose two marginal sensitivity restrictions: an odds-ratio bound Γ \Gamma for source treatment assignment and a conditional likelihood-ratio bound Λ \Lambda for source-to-target potential-outcome distribution shift. For each treatment arm and threshold y y , we derive a closed-form sharp target counterfactual CDF envelope. The envelope nests a source marginal-sensitivity map inside a target outcome-shift map, preserving two normalizations and generally improving on a single product likelihood-ratio relaxation. We prove process-level sharpness, so the envelopes are attainable as entire CDFs and can be inverted to obtain sharp target quantile bounds and sharp interval-hull QTE bounds. We then develop semiparametric theory for these nonsmooth bound processes. On regular index sets, we give the canonical gradient, including the source propensity contribution required in observational studies, and construct cross-fitted Neyman-orthogonal one-step estimators with uniform Gaussian approximation. On full index sets with active-set ties or mass points, we use Hadamard directional differentiability and subsampling-valid inference, with a primitive finite-support route for the required weak convergence. Finally, we invert simultaneous monotone CDF bands to obtain honest confidence sets for quantile and QTE interval-hull processes, and formulate the two-dimensional ( Γ , Λ ) (\Gamma,\Lambda) breakdown frontier as level-set inference for interval-hull non-refutation. Keywords: partial identification; quantile treatment effect; transportability; external validity; unmeasured confounding; marginal sensitivity model; efficient influence function; directional differentiability; breakdown frontier. 1 Introduction A common design in causal generalization combines a source study, where treatment and outcome are observed, with a target sample, where only baseline covariates are observed ( Cole and Stuart 2010 ; Stuart et al. 2011 ; Dahabreh et al. 2019 ; Pearl and Bareinboim 2011 ; Bareinboim and Pearl 2016 ; Degtiar and Rose 2023 ) . Under source internal validity and conditional transportability of potential outcomes, the target counterfactual distribution and the target quantile treatment effect are identified by standardization to the target covariate distribution ( Firpo 2007 ; Chernozhukov et al. 2013 ; Wang 2026 ) . This paper studies the harder case in which both assumptions are allowed to fail in controlled ways. The source treatment assignment may be confounded by unmeasured variables, and the conditional potential-outcome distribution in the target population may differ from the corresponding source distribution even after conditioning on observed covariates. The estimand is the target-population quantile treatment effect \Delta_{T}(\tau)=Q^{T}_{1}(\tau)-Q^{T}_{0}(\tau),\qquad\tau\in(0,1), where Q^{T}_{a} is the quantile function of the target potential outcome Y^{a} under treatment a . The target sample contains no outcomes, so even standard point identification requires transporting an entire conditional distribution from the source to the target. Sensitivity analysis for this object is not a direct extension of average-effect sensitivity analysis. Average effects are linear in the conditional potential-outcome law. QTEs require first characterizing sharp bounds on counterfactual distribution functions and then passing through a generalized inverse map, which is nonsmooth at mass points, flat regions and envelope contact sets; this feature is central in recent distributional sensitivity analysis under relaxations of unconfoundedness ( Masten et al. 2025 ) . We impose two sensitivity restrictions. First, the source population satisfies a marginal odds-ratio sensitivity model with parameter \Gamma\geq 1 : the odds of receiving treatment a after conditioning on the potential outcome Y^{a} and covariates may differ from the observed source propensity odds by at most \Gamma ( Tan 2006 ; Zhao et al. 2019 ; Dorn and Guo 2023 ; Masten et al. 2025 ) . Second, source-to-target outcome shift is controlled by a conditional likelihood-ratio bound \Lambda\geq 1 : for each treatment arm and covariate value, the target potential-outcome distribution may be an outcome-dependent tilt of the source potential-outcome distribution, with Radon–Nikodym derivative between \Lambda^{-1} and \Lambda ( Asiaee et al. 2026 ) . When (\Gamma,\Lambda)=(1,1) the standard point-identified transported QTE model is recovered. The first contribution is a closed-form sharp distributional envelope under the joint sensitivity model. Let p_{a}(y,x)=\mathbb{P}(Y\leq y\mid R=1,A=a,X=x),\qquad e_{a}(x)=\mathbb{P}(A=a\mid R=1,X=x), where R=1 denotes membership in the source sample and R=0 denotes membership in the target covariate sample. Define \ell_{\Gamma}(e)=e+(1-e)/\Gamma,\qquad u_{\Gamma}(e)=e+\Gamma(1-e), which are the lower and upper bounds on the inverse treatment-selection tilt induced by the marginal sensitivity model. The sharp source-population potential-outcome CDF bounds are C^{-}_{\Gamma}(p,e)=\max\{\ell_{\Gamma}(e)p,\,1-u_{\Gamma}(e)(1-p)\},\quad C^{+}_{\Gamma}(p,e)=\min\{u_{\Gamma}(e)p,\,1-\ell_{\Gamma}(e)(1-p)\}. The sharp transport-shift CDF maps are T^{-}_{\Lambda}(q)=\max\{\Lambda^{-1}q,\,1-\Lambda(1-q)\},\quad T^{+}_{\Lambda}(q)=\min\{\Lambda q,\,1-\Lambda^{-1}(1-q)\}. We prove that the sharp target conditional CDF envelope is b^{-}_{a,s}(y,x)=T^{-}_{\Lambda}\{C^{-}_{\Gamma}(p_{a}(y,x),e_{a}(x))\},\qquad b^{+}_{a,s}(y,x)=T^{+}_{\Lambda}\{C^{+}_{\Gamma}(p_{a}(y,x),e_{a}(x))\}, where s=(\Gamma,\Lambda) . The target marginal CDF envelope is then \psi^{-}_{a,s}(y)=\mathbb{E}\{b^{-}_{a,s}(y,X)\mid R=0\},\qquad\psi^{+}_{a,s}(y)=\mathbb{E}\{b^{+}_{a,s}(y,X)\mid R=0\}. The proof is constructive: the least-favorable source distribution is obtained by a threshold inverse-selection tilt, and the least-favorable target distribution is obtained by a second threshold outcome-shift tilt. Thus the bounds are not merely pointwise. They are attainable as entire CDF paths. Consequently the sharp target quantile bounds are q^{-}_{a,s}(\tau)=\inf\{y:\psi^{+}_{a,s}(y)\geq\tau\},\qquad q^{+}_{a,s}(\tau)=\inf\{y:\psi^{-}_{a,s}(y)\geq\tau\}, and the sharp interval-hull bounds for the target QTE are \Big[q^{-}_{1,s}(\tau)-q^{+}_{0,s}(\tau),\;q^{+}_{1,s}(\tau)-q^{-}_{0,s}(\tau)\Big]. The second contribution is efficient semiparametric theory for the sharp CDF-bound process, using tangent-space calculations for semiparametric models and orthogonal-score ideas for flexible nuisance estimation ( Tsiatis 2006 ; van der Vaart 2000 ; Chernozhukov et al. 2018 ) . At active-set regular laws the endpoint map is pathwise differentiable. If G^{-}_{s}=T^{-}_{\Lambda}\circ C^{-}_{\Gamma} or G^{+}_{s}=T^{+}_{\Lambda}\circ C^{+}_{\Gamma} denotes the active endpoint map and G_{p},G_{e} its derivatives with respect to (p,e) , the source contribution to the efficient influence function is G_{p}\{p_{a}(y,X),e_{a}(X)\}\frac{\mathbf{1}(A=a)}{e_{a}(X)}\{\mathbf{1}(Y\leq y)-p_{a}(y,X)\}+G_{e}\{p_{a}(y,X),e_{a}(X)\}\{\mathbf{1}(A=a)-e_{a}(X)\}. The second term is essential for observational source studies because the endpoint depends on the source treatment propensity through the sensitivity envelope. It disappears only when the treatment probability is known by design. Combining this conditional contribution with the target empirical distribution yields the canonical gradient. We use it to define cross-fitted one-step estimators and establish uniform asymptotic linearity over regular index sets of thresholds, treatment arms, endpoint signs and sensitivity parameters; full-index nonregular cases are handled by directional inference. The third contribution is inference that respects the nonsmooth structure of the problem, building on the delta method for quantiles and on modern inference for directionally differentiable maps ( van der Vaart 2000 ; Fang and Santos 2019 ) . The envelope maps are maxima and minima of affine functions, and the quantile map is a generalized inverse. Wald intervals for quantile endpoints are therefore fragile and require additional density and active-set assumptions. Our primary inferential object is a simultaneous confidence band for the CDF-bound process; inversion of this band yields honest confidence sets for quantile bounds and QTE bounds. Under active-set regularity the critical value can be obtained by a multiplier bootstrap for the efficient process. Without active-set separation the endpoint maps remain Hadamard directionally differentiable, and we use subsampling to obtain valid critical values for the induced non-Gaussian sup-norm limit. The fourth contribution is to treat the two-dimensional sensitivity frontier as a statistical object, in the spirit of breakdown-frontier analysis for partially identified causal models ( Masten and Poirier 2020 ; Lanners et al. 2025 ) . For a fixed quantile index \tau , define the lower and upper QTE interval-hull endpoints by \Delta^{-}_{s}(\tau) and \Delta^{+}_{s}(\tau) . The