2021
DOI: 10.1002/sim.9010
|View full text |Cite
|
Sign up to set email alerts
|

Assessment of heterogeneous treatment effect estimation accuracy via matching

Abstract: We study the assessment of the accuracy of heterogeneous treatment effect (HTE) estimation, where the HTE is not directly observable so standard computation of prediction errors is not applicable. To tackle the difficulty, we propose an assessment approach by constructing pseudo-observations of the HTE based on matching. Our contributions are three-fold: first, we introduce a novel matching distance derived from proximity scores in random forests; second, we formulate the matching problem as an average minimum… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
17
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
4
1
1

Relationship

0
6

Authors

Journals

citations
Cited by 7 publications
(17 citation statements)
references
References 37 publications
(49 reference statements)
0
17
0
Order By: Relevance
“…8 it is possible to encapsulate non-parametric functions in the logit. Such decomposition is present in the literature (Gao and Hastie, 2021) (and see Section D, and in particular Lemma 11 for details). But interpretation of all other metrics than the conditional odds ratio is clearly not obvious (see Lemma 12 in Section D) In other words, it seems that such logistic models are rather forcing all covariates and treatment to interact through the link function, preventing any simple interpretation of the causal measure (except conditional OR).…”
Section: Why Not a Logistic Model?mentioning
confidence: 95%
“…8 it is possible to encapsulate non-parametric functions in the logit. Such decomposition is present in the literature (Gao and Hastie, 2021) (and see Section D, and in particular Lemma 11 for details). But interpretation of all other metrics than the conditional odds ratio is clearly not obvious (see Lemma 12 in Section D) In other words, it seems that such logistic models are rather forcing all covariates and treatment to interact through the link function, preventing any simple interpretation of the causal measure (except conditional OR).…”
Section: Why Not a Logistic Model?mentioning
confidence: 95%
“…For each sub-challenge, the primary metric (DSS) was the difference in squared scaled BM between the nivolumab arm and chemotherapy arm, where scaled ( BM ) = 2 × ( BM − 0.5) (Table 1, Supplementary Figure 1). 32,33 Models that performed well in the nivolumab arm and randomly in the chemotherapy arm had positive primary scores. Models that performed well in the chemotherapy arm but randomly in the nivolumab arm had negative primary scores.…”
Section: Methodsmentioning
confidence: 99%
“…For each sub-challenge, the primary metric (DSS) was the difference in squared scaled BM between the nivolumab arm and chemotherapy arm, where ‫݈݀݁ܽܿݏ‬ ሺ‫ܯܤ‬ሻ ൌ 2 ൈ ‫ܯܤ‪ሺ‬‬ െ 0.5ሻ (Table 1, Supplementary Figure 1). 32,33 Models that performed well in the nivolumab arm and randomly in the chemotherapy arm had positive primary scores.…”
Section: Training and Validation Datasetsmentioning
confidence: 99%
“…The orthogonalization strategy has the distinct advantage over other methods against confounding-such as inverse propensity weighting and matching-that it is stable for extreme propensity scores and forgoes stratification. 35 Robinson 34 and Chernozhukov et al 15 use parametric models to estimate treatment effects based on residualized W and Y , but these models could be replaced by non-parametric or local parametric models 16,49 -such as model-based forests. For mean regression, Dandl et al 36 adapted the orthogonalization strategy to model-based forests.…”
Section: Model-based Forests For Observational Studiesmentioning
confidence: 99%
“…As stated above, our main goal is to assess how the orthogonalization strategy proposed for continuous outcomes could be extended to models beyond mean regression, specifically generalized linear models and transformation models. Based on Dandl et al 36 and Gao and Hastie 35 we propose two different versions of model-based forests, which should be more robust against confounding. Following Dandl et al, 36 we formulate research questions for these versions, which we aim to answer empirically in Section 4.…”
Section: Novel Model-based Forests For Observational Datamentioning
confidence: 99%