Solutions in Causal Inference
Kellyn Arnold, PhD student, School of Medicine, University of Leeds.
Do we fully understand the challenges of introducing machine learning into health research? Lessons from our (poor) understanding of linear modelling
In health research, machine learning (ML) is often hailed as the new frontier of data analytics which, combined with big data, will purportedly revolutionise delivery of healthcare and ultimately lead to more informed public health policy and clinical decision-making. However, much of the promise of ML is predicated on prediction, which is fundamentally distinct from causal inference. Nevertheless, these two concepts are often conflated in practice. I’ll briefly consider the sources and consequences of this conflation in the context of generalised linear models, as well as the potential implications for ML methods.
Peter Tennant, University Academic Fellow in Health Data Science, Leeds Institute for Data Analytics, University of Leeds.
Analyses of change: A causal inference perspective
Johannes Textor, Radboud University Medical Center, Nijmegen, The Netherlands & Visiting Associate Professor, LIDA, University of Leeds, UK.
Making DAG-based causal inference more quantitative
Directed acyclic graphs (DAGs) are sequential, discrete-time models data-generating processes. DAGs have become popular due to the principled and elegant fashion in which they allow to formalize causal inference. In model-based research, much attention is usually paid to the aspect of model checking -- testing whether a model actually fits the data it is supposed to describe. Specifically, models are often only considered credible after they have survived serious falsification attempts. Surprisingly, DAG models are rarely ever tested at present, and at least in Epidemiology there even appears to be a widely-held belief that DAG models should not be tested at all. In this talk, I will make the case that DAG model checking should be an integrated part of any DAG-based causal inference. I will explain how DAGs can be tested by means of the conditional independencies that they imply, how this can be done in practice with continuous and/or categorical data, which problems we are likely to encounter, and how to address these problems. I will also show how DAG testing methodology can be converted into a means to reason about the strengths of certain hypothesized pathways in a DAG.
Видео Solutions in Causal Inference канала RoyalStatSoc
Do we fully understand the challenges of introducing machine learning into health research? Lessons from our (poor) understanding of linear modelling
In health research, machine learning (ML) is often hailed as the new frontier of data analytics which, combined with big data, will purportedly revolutionise delivery of healthcare and ultimately lead to more informed public health policy and clinical decision-making. However, much of the promise of ML is predicated on prediction, which is fundamentally distinct from causal inference. Nevertheless, these two concepts are often conflated in practice. I’ll briefly consider the sources and consequences of this conflation in the context of generalised linear models, as well as the potential implications for ML methods.
Peter Tennant, University Academic Fellow in Health Data Science, Leeds Institute for Data Analytics, University of Leeds.
Analyses of change: A causal inference perspective
Johannes Textor, Radboud University Medical Center, Nijmegen, The Netherlands & Visiting Associate Professor, LIDA, University of Leeds, UK.
Making DAG-based causal inference more quantitative
Directed acyclic graphs (DAGs) are sequential, discrete-time models data-generating processes. DAGs have become popular due to the principled and elegant fashion in which they allow to formalize causal inference. In model-based research, much attention is usually paid to the aspect of model checking -- testing whether a model actually fits the data it is supposed to describe. Specifically, models are often only considered credible after they have survived serious falsification attempts. Surprisingly, DAG models are rarely ever tested at present, and at least in Epidemiology there even appears to be a widely-held belief that DAG models should not be tested at all. In this talk, I will make the case that DAG model checking should be an integrated part of any DAG-based causal inference. I will explain how DAGs can be tested by means of the conditional independencies that they imply, how this can be done in practice with continuous and/or categorical data, which problems we are likely to encounter, and how to address these problems. I will also show how DAG testing methodology can be converted into a means to reason about the strengths of certain hypothesized pathways in a DAG.
Видео Solutions in Causal Inference канала RoyalStatSoc
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Building and Using DAGs for Causal InferenceLearning Representations Using Causal Invariance - Leon BottouuseR! 2020: Causal inference in R (Lucy D'Agostino McGowan, Malcom Barrett), tutorialMichael Johns: Propensity Score Matching: A Non-experimental Approach to Causal... | PyData NYC 2019All of causal discovery - Frederick EberhardtStatistical vs. Causal Inference: Causal Inference BootcampJudea Pearl, 2012 ACM A.M. Turing Award Lecture "The Mechanization of Causal Inference"Causal NetworksCRITICAL THINKING - Fundamentals: Correlation and CausationMachine learning for causal inference: Magic elixir or fool’s gold?Causation I 1 ReverseIntermittent Fasting & Hunger - What the Science saysDesign of the Oregon Health Experiment: Causal Inference BootcampJuliaCon 2018 | OMEGA: Fast, causal Inference from Simple Parts | Zenna TavaresChoosing a Statistical TestSusan Athey Guest Talk - Estimating Heterogeneous Treatment EffectsThe Common Trends Assumption: Causal Inference BootcampPython and the Holy Grail of Causal Inference - Dennis Ramondt, Huib KeeminkCS 201 JUDEA PEARL MARCH 9 2021Bayesian networks and causality by Richard Neapolitan