- 20
- 41 231
Putnam Data Sciences
United States
เข้าร่วมเมื่อ 22 ม.ค. 2020
Webinars and Tutorials on Targeted Learning, Causal Inference, Machine Learning. Slides can be downloaded from www.putnamds.com/#TargetedLearning .
Funding was received from the US Food and Drug Administration, as part of a project titled, "A Targeted Learning Framework for Causal Effect Estimation Using Real World Data," (FDA CONTRACT 75F40119C10155). These videos are available to the public at no cost. Content reflects the views of the presenter, and should not be construed to represent FDA's views or policies.
Funding was received from the US Food and Drug Administration, as part of a project titled, "A Targeted Learning Framework for Causal Effect Estimation Using Real World Data," (FDA CONTRACT 75F40119C10155). These videos are available to the public at no cost. Content reflects the views of the presenter, and should not be construed to represent FDA's views or policies.
What is an Influence Curve?
This is the first of two Targeted Learning Briefs concerning the influence curve. The influence curve, also called the influence function, was introduced in 1974 by Frank Hampel, as an aid for studying the robustness of an estimator in the face of small changes in the distribution of the data. In this Brief we'll go over Hampel's example of the influence curve of an estimator of the mean. (F. Hampel. The Influence Curve and Its Role in Robust Estimation. Journal of the American Statistical Association,1974: 69(346), p383-393)
มุมมอง: 1 479
วีดีโอ
The Ratio of Bias to Standard Error: The secret behind poor confidence interval coverage
มุมมอง 223ปีที่แล้ว
This Targeted Learning Brief discusses how the ratio of bias to standard error affects confidence interval coverage. As sample size increases, confidence intervals grow narrower and narrower. In this era of Big Data, confidence intervals can be quite small. Although that sounds like a good thing, we'll see that if you have a biased estimator with a small standard error, you can easily be misled...
Why V-Fold Cross Validation?
มุมมอง 145ปีที่แล้ว
In this Targeted Learning Brief we'll learn what cross-validation is used for, and the benefits of a particular type called "V-fold" cross-validation. This project funded by FDA Contract 75F40119C10155, A Targeted Learning Framework for Causal Effect Estimation Using Real World Data. Available for public viewing at no cost. The content of this video are those of the presenter(s) and do not nece...
An Overview of Targeted Learning
มุมมอง 809ปีที่แล้ว
Targeted learning briefs are a series of instructional videos on causal inference machine learning and targeted learning. This first brief provides an overview of targeted learning.
Targeted Learning: Towards a future informed by real world evidence
มุมมอง 1K3 ปีที่แล้ว
Presented by Dr. Mark van der Laan, Professor of Statistics and Biostatistics at University of California, Berkeley. Learning from data to support regulatory decision making has to do with translating a real world data experiment into a statistical estimation problem, providing valid inference and assessing the validity of a causal interpretation of the study finding. Producing real world evide...
BAA Project to Advance Regulatory Science&Leverage Real World Evidence in Regulatory Decision Making
มุมมอง 3073 ปีที่แล้ว
Co-presented by Dr. John Concato, MD, MS, MPH, Associate Director of the Office of Medical Policy for Real-Word Evidence Analytics in the Center for Drug Evaluation and Research (CDER), FDA, and Dr. Hana Lee, PhD, a Senior Statistical Reviewer of the Office of Biostatistics in the CDER, FDA.
Expert Augmented Machine Learning
มุมมอง 3893 ปีที่แล้ว
Presented by Dr. Gilmer Valdes, assistant professor with dual appointment in the departments of Radiation Oncology and Epidemiology and Biostatics at University of California, San Francisco. The talk is based on recently published work: Gennatas, E. D., et al. Expert-augmented machine learning. March 3, 2020 117 (9) 4571-4577. Proceedings of the National Academy of Sciences. www.pnas.org/conten...
Challenges and Solutions in the Analysis of Cluster Randomized Trials
มุมมอง 1.2K3 ปีที่แล้ว
Presented by Laura Balzer, Assistant Professor of Biostatistics and Director of the Causality Lab at UMass Amherst, based on a recent publication, arxiv.org/abs/2106.15737 . Cluster randomized trials (CRTs) randomly assign an intervention to groups of individuals (e.g., clinics or communities), and measure outcomes on individuals in those groups. While offering many advantages, this experimenta...
Practical Considerations for Specifying a Super Learner
มุมมอง 1.1K3 ปีที่แล้ว
Presented by Rachael Phillips. Super learner is a machine learning algorithm that uses cross-validation and a loss function to optimally combine a diversity of candidate algorithms into a “super learner” ensemble. The super learner is extremely flexible and entirely pre-specified, but how should it be constructed? In this webinar, we provide a flowchart for constructing a super learner in pract...
Developing a Targeted Learning-Based Statistical Analysis Plan
มุมมอง 6633 ปีที่แล้ว
Presented by Susan Gruber. When the goal is to estimate causal effects from data it is crucial to pre-specify the entire data analysis. This is especially important when using data-adaptive statistical methodology, in order to ensure interpretable results are obtained from a valid, reproducible analyses. We’ll consider a sample Targeted Learning (TL)-based statistical analysis plan that illustr...
Targeted Machine Learning in Action in the ICU
มุมมอง 3863 ปีที่แล้ว
Dr. Romain Pirracchio, Chief of Anesthesia and Properative Medicine at Zuckerberg San Fracisco General Hospital and Trauma Center, discusses using Targeted Learning to transform a curative heatlh care model into a preventative service.
Highly Adaptive Lasso (HAL) in Causal Inference
มุมมอง 1.7K4 ปีที่แล้ว
Dr. Mark van der Laan introduces the Highly Adaptive Lasso, a novel nonparametric (maximum likelihood) estimator of regression functions that makes global smoothness assumptions which are exceedingly mild but nevertheless sufficient to achieve a remarkably fast convergence rate. Incorporating HAL to estimate conditional probability distributions that are an intrinsic component of causal inferen...
Cross-validated Targeted Maximum Likelihood Estimation (CV-TMLE)
มุมมอง 1.4K4 ปีที่แล้ว
Alan Hubbard is a Professor of Biostatistics at UC Berkeley. His research focuses on the application of statistics to population studies with particular expertise in semi-parametric models and the use of machine learning in causal inference, as well as applications in high dimensional biology.
Covariate adjustment in randomized studies with time-to-event endpoints
มุมมอง 1.1K4 ปีที่แล้ว
Dr. Iván Díaz, Assistant Professor of Biostatistics at Weill Cornell Medicine, works on methods for statistical learning and causal inference from observational and randomized studies with complex designs and datasets. He discusses the advantages of covariate adjustment in randomized studies with time-to-event endpoints, and application of TMLE and Super Learning.
Practical Issues in Targeted Learning
มุมมอง 8224 ปีที่แล้ว
Dr. David Benkeser, Assistant Professor in the Department of Biostatistics and Bioinformatics at Emory University's Rollins School of Public Health discusses real world considerations when estimating causal effects, and considerations when implementing Targeted Learning.
Causal Inference of Longitudinal Exposures, presented by Dr. Mireille Schnitzer
มุมมอง 2.9K4 ปีที่แล้ว
Causal Inference of Longitudinal Exposures, presented by Dr. Mireille Schnitzer
Introduction to Bayesian Additive Regression Trees (BART) for Causal Inference
มุมมอง 11K4 ปีที่แล้ว
Introduction to Bayesian Additive Regression Trees (BART) for Causal Inference
2. An Introduction toTargeted Maximum Likelihood Estimation of Causal Effects
มุมมอง 6K4 ปีที่แล้ว
2. An Introduction toTargeted Maximum Likelihood Estimation of Causal Effects
1. Targeted Machine Learning for Causal Inference based on Real World Data
มุมมอง 3.6K4 ปีที่แล้ว
1. Targeted Machine Learning for Causal Inference based on Real World Data
Clear explanations! No heavy notations. Thank you for the lecture.
loved the video, really hope part II comes soon :)
excellent!!
Thanks for making this talk public - a great resource. Interested in whether you might be able to provide a reference for the sensitivity method presented around 37:50. Is this similar to the approach presented in Diaz and van der Laan (2013)?
Yes, exactly. We describe it in practical terms in several recent publications: a paper about how TL can inform writing an SAP paper is currently under review, but see also arxiv.org/abs/2205.08643 , "Targeted Learning: Towards a future informed by RWE"
@@putnamdatasciences1309 Fantastic - thanks for confirming and many thanks for the quick response, greatly appreciated. Will certainly check out the linked article.
very (very) good stuff. thank ye
As for those model comparisons (bias/RMSE), I assume these stats are based on the training set? If so, than I would rather see the RMSE stats on a hold-out test set as a fair comparison for robustness.
The RMSE was with respect to estimating a marginal causal effect across thousands of datasets. It had nothing to do with the accuracy of the outcome predictions. For details on the 2019 ACIC Data Challenge see sites.google.com/view/acic2019datachallenge/home
@@putnamdatasciences1309 Thank you for providing this context. I like the thought process behind this ML algo, so I will definitely run some tests.
this is a great presentation - thanks for uploading!
Thank you so much!
39:10 for list of references