Subgroup Analysis via Recursive Partitioning

Su, Xiaogang; Tsai, Chih‐Ling; Wang, Hansheng; Nickerson, David M.; Li, Bogong

doi:10.2139/ssrn.1341380

Cited by 247 publications

(406 citation statements)

References 20 publications

Supporting

Mentioning

404

Contrasting

Order By: Relevance

“…(Instead of personalizing predictions for each individual, this approach only provides treatment effect estimates for leaf-wise subgroups whose size must grow to infinity.) Other related approaches include those of Su et al [2009] and Zeileis et al [2008], which build a tree for treatment effects in subgroups and use statistical tests to determine splits; however, these papers do not analyze bias or consistency properties.…”

Section: Related Workmentioning

confidence: 99%

Estimation and Inference of Heterogeneous Treatment Effects using Random Forests

Wager

Athey

2018

Journal of the American Statistical Association

2,015

1,780

View full text Add to dashboard Cite

Many scientific and engineering challenges-ranging from personalized medicine to customized marketing recommendations-require an understanding of treatment effect heterogeneity. In this paper, we develop a non-parametric causal forest for estimating heterogeneous treatment effects that extends Breiman's widely used random forest algorithm. In the potential outcomes framework with unconfoundedness, we show that causal forests are pointwise consistent for the true treatment effect, and have an asymptotically Gaussian and centered sampling distribution. We also discuss a practical method for constructing asymptotic confidence intervals for the true treatment effect that are centered at the causal forest estimates. Our theoretical results rely on a generic Gaussian theory for a large family of random forest algorithms. To our knowledge, this is the first set of results that allows any type of random forest, including classification and regression forests, to be used for provably valid statistical inference. In experiments, we find causal forests to be substantially more powerful than classical methods based on nearest-neighbor matching, especially in the presence of irrelevant covariates.

show abstract

Section: Related Workmentioning

confidence: 99%

Estimation and Inference of Heterogeneous Treatment Effects using Random Forests

Wager

Athey

2018

Journal of the American Statistical Association

2,015

1,780

View full text Add to dashboard Cite

show abstract

“…Develop new and novel methods to identify multiple participant characteristics or clusters of moderators that would identify who is most or least likely to benefit. [94][95][96] 2. To apply individual participant data meta-analysis to homogenous pooled datasets as this would improve statistical power.…”

Section: Summary Of Reviewsmentioning

confidence: 99%

“…They are the interaction tree (IT) and subgroup identification based on a differential effect search (SIDES). 94,96 These methods were initially developed and implemented in a single trial setting. Therefore, they have to be extended so that they can be applied in an individual participant data (IPD) meta-analyses framework.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Identifying back pain subgroups: developing and applying approaches using individual patient data collected within clinical trials

Patel

Hee

Mistry

et al. 2016

Programme Grants Appl Res

View full text Add to dashboard Cite

Total word count including appendix: 93,499 ii ABSTRACT BackgroundThere is good evidence that therapist delivered interventions have modest beneficial effects for people with low back pain (LBP). Identification of subgroups of people with LBP who may benefit from these different treatment approaches is an important research priority. Aim and objectivesOverall aim was to improve the clinical and cost-effectiveness of LBP treatment by providing patients, their clinical advisors, and health service purchasers with better information about which participants are most likely to benefit from which treatment To achieve these objectives required substantial methodological work including the development and evaluation of some novel statistical approaches. This programme of work was not designed to analyse main effect of interventions and no such interpretations should be made.iii MethodsFirstly, we reviewed the literature on treatment moderators and subgroups. We initially invited investigators of trials of therapist-delivered interventions for LBP with >179 participants to share their data with us; some further smaller trials offered to us were also included. Using these trials we developed a repository of individual participant data of therapist delivered interventions for LBP. Using this dataset we sought to identify which participant characteristics, if any, predict response to different treatments (moderators) for clinical and cost effectiveness outcomes.We did an ANCOVA to identify potential moderators to apply in our main analyses.Subsequently we developed and applied three methods of subgroup identification; recursive partitioning (interaction trees and subgroup identification based on a differential effect search), adaptive risk group refinement, and an individual participant data indirect network meta-analysis to identify sub-groups defined by multiple parameters. ResultsWe included data from 19 randomised controlled trials with 9,328 participants (mean age 49 years, 57% females). Our prespecified analyses using recursive partitioning and adaptive risk group refinement performed well and allowed us to identify some subgroups. The differences in the effect size in the different subgroups were typically small, and unlikely to be clinically meaningful. Increasing baseline severity on the outcome of interest was the strongest driver of sub-group identification that we identified. Additionally we explored the application of Bayesian indirect network metaanalysis. This method produced varying probabilities that a particular treatment choice would be most likely to be effective for a specific patient profile. ConclusionThese data lack clinical or cost-effectiveness justification for the use of baseline characteristics in the development of subgroups for back pain. The methodological developments from this work have the potential to be applied in other clinical areas.iv The pooled repository database will serve as a valuable resource to the LBP research community. FundingFunding from the NIHR Programme Grants for Ap...

show abstract

“…The reason why we compare our sHinge method with the ROWSi is that Xu et al (2015) showed by simulation that the ROWSi was superior over the method solving (9) with h k replaced by the hinge loss, which was proposed in Zhao et al (2012) except that LASSO penalty instead of L 2 penalty was used for variable selection. Xu et al (2015) also showed by simulation that ROWSi was superior over other four recently proposed methods, the interaction tree by Su, Tsai, Wang, Nickerson, and Li (2009), the virtual twins by Foster, Taylor, and Ruberg (2011), the logistic regression with LASSO penalty by Qian and Murphy (2011), and the FindIt by Imai and Ratkovic (2013).…”

Section: Simulation Resultsmentioning

confidence: 89%

Personalised treatment assignment maximising expected benefit with smooth hinge loss

Liu

Shao

2017

Statistical Theory and Related Fields

View full text Add to dashboard Cite

In personalised medicine, the goal is to make a treatment recommendation for each patient with a given set of covariates to maximise the treatment benefit measured by patient's response to the treatment. In application, such a treatment assignment rule is constructed using a sample training data consisting of patients' responses and covariates. Instead of modelling responses using treatments and covariates, an alternative approach is maximising a response-weighted target function whose value directly reflects the effectiveness of treatment assignments. Since the target function involves a loss function, efforts have been made recently on the choice of the loss function to ensure a computationally feasible and theoretically sound solution. We propose to use a smooth hinge loss function so that the target function is convex and differentiable, which possesses good asymptotic properties and numerical advantages. To further simplify the computation and interpretability, we focus on the rules that are linear functions of covariates and discuss their asymptotic properties. We also examine the performances of our method with simulation studies and real data analysis.

show abstract

Subgroup Analysis via Recursive Partitioning

Cited by 247 publications

References 20 publications

Estimation and Inference of Heterogeneous Treatment Effects using Random Forests

Estimation and Inference of Heterogeneous Treatment Effects using Random Forests

Identifying back pain subgroups: developing and applying approaches using individual patient data collected within clinical trials

Personalised treatment assignment maximising expected benefit with smooth hinge loss

Contact Info

Product

Resources

About