Dear colleague in causality research,<\/p>\n
This is an End-of-Winter Greetings from the\u00a0UCLA Causality blog, welcoming you back\u00a0to a spring-time discussion in causality-related issues.<\/p>\n
This message contains
\n1. Topics under discussion
\n2. New results
\n3. Information on courses, lectures, and conferences.<\/p>\n
1. Discussions inviting comments<\/strong><\/span><\/p>\n 1.1.\u00a0“Principal Stratification – A goal or a tool?” If you have comments, ideas or objections, you are\u00a0invited to communicate them to the IJB’s \u00a0Editor,\u00a0“Nicholas P. Jewell” <mm-11332-3261687@bepress.com<\/a>>\u00a0or\/and, if you wish, cross-post them on this blog.<\/p>\n 1.2. “Comments and Controversies: Graphical models,\u00a0potential outcomes and causal inference: Comment on\u00a0Lindquist and Sobel” This note comments on a paper published in NeuroImage\u00a0which argues (yes, again) that the potential outcome model is\u00a0somehow superior, more rigorous or more principled than the\u00a0structural models used in fMRI research.\u00a0To further illuminate the logic of such claims I\u00a0have added a section (4.4.2) in this paper: 1.3.\u00a0“The Causal Mediation Formula – A practitioner guide to the\u00a0assessment of causal pathways” This paper present mediation analysis\u00a0to researchers in the tradition of Baron and Keney\u00a0(1986), and shows through examples\u00a0how “the percentage explained by mediation”\u00a0and “the percentage owed to mediation” are estimated in\u00a0nonlinear models with both continuous and categorical variables.<\/p>\n 1.4.\u00a0Simpson’s Paradox 1.5.\u00a0“The ETT Paradox \u00a0(or, the curse of free will)” 2. New Results<\/strong><\/span><\/p>\n 2.1.\u00a0A newly posted paper, “Controlling Selection Bias in Causal Inference”\u00a0http:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r381.pdf<\/a> gives graphical and algebraic conditions for the\u00a0removal of selection bias and the recovery of\u00a0covariate-specific effect measures.<\/p>\n 2.2.\u00a0A new section (Section 5) in\u00a0http:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r372.pdf<\/a> generalizes the concept of transportability\u00a0from experimental to observational studies,and\u00a0shows how one can avoid re-learning things\u00a0from scratch when moving to a new population, new domain,\u00a0or a new environment.<\/p>\n 2.3.\u00a0After months of struggling with the literature\u00a0of “surrogate endpoints” we feel that we now have a\u00a0fairly satisfactory theory of surrogacy.\u00a0It is based on the idea that a surrogate\u00a0should serve not merely as a good predictor of outcomes,\u00a0but also as ROBUST predictor of effects in the face of changing\u00a0external conditions. \u00a0See section 6 in\u00a0http:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r372.pdf<\/a><\/p>\n 3.\u00a0Courses, Lectures and Conferences<\/strong><\/span><\/p>\n 3.1.\u00a0Causal Inference Course 3.2.\u00a02011 Atlantic Causal Conference 3.3.\u00a0Errata for Causality (2010)
\nhttp:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r382.pdf<\/a>
\nPosted for discussion by the International Journal of\u00a0Biostatistics (IJB), this paper questions whether\u00a0studies based on Principal Stratification\u00a0target quantities that researchers truly care about.<\/p>\n
\nhttp:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r380.pdf<\/a><\/p>\n
\nhttp:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r370.pdf<\/a>
\nwhich demonstrates how potential outcomes can be generated, on\u00a0demand, from a simple structural model, and no one can\u00a0tell where they came from.\u00a0Enjoy.<\/p>\n
\nhttp:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r379.pdf<\/a><\/p>\n
\nSander Greenland brought to my attention a recent paper\u00a0in Synthese (Sept. 28, 2010) claiming that Simpson’s paradox is NOT\u00a0rooted in causal, but in some other kind of illusion.\u00a0I remain convinced of the former, and have accordingly\u00a0modified the Simpson Paradox entry in Wikipedia to\u00a0reinforce the causal illusion theory. You might wish\u00a0to add your take on the subject.<\/p>\n
\nThis paradox would be appreciated by those who\u00a0are fascinated, like me, by our ability to determine, from data\u00a0alone, if one would have been better off acting differently than\u00a0one actually did. This can lead to a cycle of inevitable regret,\u00a0and provokes some naughty thoughts.
\nhttp:\/\/ftp.cs.ucla.edu\/pub\/stat_ser\/r375.pdf<\/a><\/p>\n
\nThomas Richardson and Michael Hudgens are once again\u00a0teaching Causal Inference\u00a0June 13-15, 2011 in the Summer Institute here at U\u00a0Washington. They have funds to support tuition waivers and some travel for\u00a0students and postdocs. The website is\u00a0http:\/\/depts.washington.edu\/sismid\/<\/a><\/p>\n
\nThe 2011 Atlantic Causal Conference will take place at\u00a0the University of Michigan School of Public Health in Ann\u00a0Arbor, Michigan, Thursday May 19th and Friday May 20th.\u00a0See
\nhttp:\/\/www.sph.umich.edu\/biostat\/2011acic\/index.html<\/a> for
\nContact: Mike Elliott at\u00a0mrelliot@umich.edu<\/a> or Ben Hansen at\u00a0ben.hansen@umich.edu<\/a><\/p>\n
\nFYI, Cambridge University Press has come up with a\u00a0new printing of my book Causality, which corrects a few\u00a0errors in the 2009 edition. Please advise\u00a0students to insist on a copy saying “reprinted 2010”.
\nIf you have an older copy, you can find the corrections\u00a0marked in red here:
\nhttp:\/\/bayes.cs.ucla.edu\/BOOK-09\/errata_scanned_pages7-28-10.pdf<\/a><\/p>\n