~/talk

An Introduction to Potential Outcomes, DAGs and Single-World Intervention Graphs

Talkat Causality Boot Camp · 2022-01-18..01-22· [machine-learning]
links:none on file
slides:none on file

Presents a complete, constructive identification procedure: re-express P(Y|do(X)) as a sum of district-specific intervention kernels, then test each kernel for identifiability by iteratively applying a graph operation called the fixing operation that cuts incoming arrows and reweights kernels. Delivers a compact one-line ID algorithm (ancestors-of-Y after removing X-edges, check that all districts are intrinsic sets), shows worked front-door and sequential-randomization examples, and gives a simple hedge-based criterion proving nonidentifiability when the test fails.

AI-written summary · about

  1. 0:00Start
  2. 2:54General Identification Problem
  3. 3:13Intervention Distributions
  4. 3:26Intervention Distribution
  5. 11:43Identification Question
  6. 13:41Latent Projection
  7. 17:46Lane Projection
  8. 20:38Notes of Ancestral Descendants
  9. 24:15Fixed and Random Nodes
  10. 25:07Schmitzer Pearl Algorithm
  11. 27:12Decompose the Query
  12. 35:05Fixable Vertices
  13. 38:11Examples
  14. 51:10One Line Id Algorithm
  15. 55:13Nested Markov Model
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.