(For the ten or fifteen people who care about Pearl’s do-calculus):
There’s a new paper on arXiv (54 pages) claiming a graphical method for determining when the effect of an intervention (on a single variable) is or is not identifiable. I’ve not read the paper in detail, but it is fairly readable (relative to the typical research paper).
Reading that paper led me to discover that the author has written an introduction to the do-calculus (16 pages) which is also quite readable. Lots of nice pictures for wrapping one’s head around d-separation and the like.
(For the ten or fifteen people who care about Pearl’s do-calculus):
There’s a new paper on arXiv (54 pages) claiming a graphical method for determining when the effect of an intervention (on a single variable) is or is not identifiable. I’ve not read the paper in detail, but it is fairly readable (relative to the typical research paper).
Reading that paper led me to discover that the author has written an introduction to the do-calculus (16 pages) which is also quite readable. Lots of nice pictures for wrapping one’s head around d-separation and the like.