[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-290671-105":59,"doc-detail-290671-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","on-doubly-robust-inference-for-double-machine-learning-in-semiparametric-regression","On doubly robust inference for double machine learning in semiparametric regression","","Machine learning is increasingly used to adjust for confounding when estimating causal effects, but plug-in bias can make conventional exposure-effect p-values and confidence intervals unreliable. The paper builds on targeted maximum likelihood estimation and double machine learning, which learn conditional means of both outcome and exposure and require sufficiently fast convergence of predictions. For partially linear regression models, it presents an implementation producing doubly robust exposure-effect estimators with small bias even if one first-stage prediction does not converge. Simulations indicate reduced bias and improved confidence-interval coverage in moderate samples.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/on-doubly-robust-inference-for-double-machine-learning-in-semiparametric-regression/290671/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/on-doubly-robust-inference-for-double-machine-learning-in-semiparametric-regression/290671.png","ImageObject",300,407,{"name":92,"@type":93},"Levi","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-26","2026-09-17",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",12,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"Why do plug-in bias issues arise in machine-learning-based causal effect estimation?","Question",{"text":113,"@type":114},"Exposure-effect estimators that plug in machine-learning predictions can produce plug-in bias, which undermines naive p-values and confidence intervals.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What does double machine learning require for valid inference?",{"text":118,"@type":114},"Valid inference is obtained when the conditional-mean predictions for both outcome and exposure converge sufficiently fast to the truth.",{"name":120,"@type":111,"acceptedAnswer":121},"How does the paper’s proposed approach achieve double robustness?",{"text":122,"@type":114},"For partially linear regression models, it constructs exposure-effect estimators whose tests and confidence intervals remain doubly robust, yielding small bias even when one first-stage prediction fails to converge.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},290671,1789642338,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":130,"read_time":145,"preview_page_count":4},7971461740909,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","View metadata, citation and similar [papers at ](papers at core.ac.uk)[core.ac.uk](papers at core.ac.uk) brought to you by CORE  \n[provided by](provided by arXiv.org)[ arXiv.org](provided by arXiv.org) e-Print Archive  \narXiv :2 107 .06124v1 [ stat .ME] 13 Jul 2021  \nOn doubly robust inference for double machine learning  \nOliver Dukes 1 , Stijn Vansteelandt 1 ;2 and David Whitney 3  \n1 : Ghent University, Ghent, Belgium  \n2 : London School of Hygiene and Tropical Medicine, London, U.K.  \n3 : Imperial College London, London, U.K.  \nJuly 14, 2021  \nAbstract  \nDue to concerns about parametric model misspeci􀀌cation, there is interest in using machine learning to adjust for confounding when evaluating the causal effect of an exposure on an outcome. Unfortunately, exposure e􀀋ect estimators that rely on machine learning predictions are generally subject to so-called plug-in bias, which can render naive p-values and con􀀌dence intervals invalid. Progress has been made via proposals like targeted maximum likelihood estimation and more recently double machine learning, which rely on learning the conditional mean of both the outcome and exposure. Valid inference can then be obtained so long as both predictions converge (su􀀎ciently fast) to the truth. Focusing on partially linear regression models, we show that a speci􀀌c implementation of the machine learning techniques can yield exposure e􀀋ect estimators that have small bias even when one of the 􀀌rststage predictions does not converge to the truth. The resulting tests and con􀀌dence intervals are doubly robust. We also show that the proposed estimators may fail tobe regular when only one nuisance parameter is consistently estimated; nevertheless, we observe in simulation studies that our proposal leads to reduced bias and improved con􀀌dence interval coverage in moderate samples.  \nKey words: causal inference, doubly robust estimation, semiparametric inference.  \n1 Introduction  \nIn recent years, there has been a resurgence of interest in so-called doubly robust estimators (Scharfstein et al., 1999; Robins and Rotnitzky, 2001) . These estimators require  \nestimation of two separate nuisance parameters, and are consistent so long as one of the nuisance parameters is consistently estimated. When these estimators were introduced, focus was initially on postulating parametric working models, where at least oneof the models should be correctly speci􀀌ed (Bang and Robins, 2005) . However, it turns out that doubly robust methods are additionally advantageous when the nuisance parameters are estimated nonparametrically; for example, using modern machine learning techniques. If both are estimated consistently, then under some additional conditions, standard parametric-rate inference can be performed on the target parameter based on a simple sandwich estimator or the non-parametric bootstrap. This is in spite of thenonparametric estimators having a potentially large bias and complex distribution. Such ideas underpin developments in targeted maximum likelihood estimation (TMLE) (van der Laan and Rose, 2011), and double machine learning (DML) (Chernozhukov et al., 2018) . Doubly robust estimators are a subclass of the methods based on orthogonal estimating functions considered by Chernozhukov et al. (2018); an advantage of this particular subclass is that parametric-rate inference for the target parameter can be achieved even when one nuisance parameter is estimated at a slow rate, so long as this is compensated by fast estimation of the other nuisance parameter.  \nHowever, these nice properties no longer hold when one of the nuisance parameters isnot estimated consistently. Although machine learning methods are more 􀀍exible than traditional parametric estimators, they may still be guilty of extrapolation in causal inference and missing data problems e.g. where treated and untreated patients are very di􀀋erent in terms of the distribution of their covariates. Although the doubly robust estimator remains","cbCaipGaYLeCk98S","https://ap.wps.com/l/cbCaipGaYLeCk98S","pdf",488302,38,"English","# Introduction\n## Doubly robust estimators and nuisance parameters\n## Parametric-rate inference with machine learning\n## Failures under inconsistent nuisance estimation\n## Related work: TMLE and doubly robust hypothesis testing","[{\"question\":\"Why do plug-in bias issues arise in machine-learning-based causal effect estimation?\",\"answer\":\"Exposure-effect estimators that plug in machine-learning predictions can produce plug-in bias, which undermines naive p-values and confidence intervals.\"},{\"question\":\"What does double machine learning require for valid inference?\",\"answer\":\"Valid inference is obtained when the conditional-mean predictions for both outcome and exposure converge sufficiently fast to the truth.\"},{\"question\":\"How does the paper’s proposed approach achieve double robustness?\",\"answer\":\"For partially linear regression models, it constructs exposure-effect estimators whose tests and confidence intervals remain doubly robust, yielding small bias even when one first-stage prediction fails to converge.\"}]","On doubly robust inference for double machine learning in semiparametric regression | PDF",96]