Day 3: Search Continued
|
|
- Jason Mason
- 5 years ago
- Views:
Transcription
1 Center for Causal Discovery Day 3: Search Continued June 15, 2015 Carnegie Mellon University 1
2 Outline Models Data 1) Bridge Principles: Markov Axiom and D-separation 2) Model Equivalence 3) Model Search A. For Patterns B. For PAGs 4) Multiple Regression vs. Model Search 5) Measurement Issues and Latent Variables 2
3 Search Results? 1) Charitable Giving 2) Lead and IQ 3) Timberlake and Williams 3
4 Constraint-based Search for Patterns 1) Adjacency phase 2) Orientation phase 4
5 Constraint-based Search for Patterns: Adjacency phase X and Y are not adjacent if they are independent conditional on any subset that doesn t X and Y 1) Adjacency Begin with a fully connected undirected graph Remove adjacency X-Y if X _ _ Y any set S
6 X1 X2 X3 Causal Graph X4 Inde pendcie s X1 X2 X1 X4 {X3} X2 X4 {X3} Begin with: X1 X3 X4 X2 From X1 X1 X2 X3 X4 X2 From X1 X1 X4 {X3} X3 X4 X2 From X2 X4 {X3} X1 X3 X4 X2
7 Constraint-based Search for Patterns: Orientation phase 2) Orientation Collider test: Find triples X Y Z, orient according to whether the set that separated X-Z contains Y Away from collider test: Find triples X Y Z, orient Y Z connection via collider test Repeat until no further orientations Apply Meek Rules 7
8 Search: Orientation Patterns Y Unshielded X Y Z Test: X _ _ Z S, is Y S No Yes Collider X Y Z Non-Collider X Y Z X Y Z X Y Z X Y Z
9 Search: Orientation Away from Collider Test Conditions X 1 X 2 X 3 1) X 1 - X 2 adjacent, and into X 2. 2) X 2 - X 3 adjacent 3) X 1 - X 3 not adjacent Test No X 1 _ _ X 3 S, X 2 S Yes X 1 X 3 X 1 X 3 X 2 X 2
10 Search: Orientation After Adjacency Phase X 1 X 3 X 4 X 2 Collider Test: X1 X3 X2 X 1 X1 _ _ X2 X 3 X 4 X 2 Away from Collider Test: X1 X3 X4 X2 X3 X4 X 1 X1 _ _ X4 X3 X 3 X 4 X2 _ _ X4 X3 X 2
11 Away from Collider Power! X 1 X 2 X 3 X 1 _ _ X 3 S, X 2 S X 1 X 2 X 3 X 2 X 3 oriented as X 2 X 3 Why does this test also show that X 2 and X 3 are not confounded? X 1 X 2 X 3 C X 1 X 2 X 3 X 1 _ _ X 3 S, X 2 S X 1 _ _ X 3 S, X 2 S, C S
12 Independence Equivalence Classes: Patterns & PAGs Patterns (Verma and Pearl, 1990): graphical representation of d-separation equivalence among models with no latent common causes PAGs: (Richardson 1994) graphical representation of a d- separation equivalence class that includes models with latent common causes and sample selection bias that are d-separation equivalent over a set of measured variables X 12
13 PAGs: Partial Ancestral Graphs PAG X 1 X 2 X 3 Represents X 2 X 1 X 1 X 2 T 1 T 1 X 3 X 3 X 1 X 2 X 1 X 2 etc. X 3 T 1 T 2 X 3 13
14 PAGs: Partial Ancestral Graphs Z 1 Z 2 PAG X Y Represents Z 1 Z 2 Z 1 Z 2 T 1 X 3 Y X 3 Y etc. T 1 Z 1 Z 2 Z 1 Z 2 T 1 T 2 X 3 Y X 3 Y 14
15 PAGs: Partial Ancestral Graphs What PAG edges mean. X 1 X 2 X 1 and X 2 are not adjacent X 1 X 2 X 2 is not an ancestor of X 1 X 1 X 2 No set d-separates X 2 and X 1 X 1 X 2 X 1 is a cause of X 2 X 1 X 2 There is a latent common cause of X 1 and X 2 15
16 PAG Search: Orientation PAGs Y Unshielded X Y Z X _ _ Z Y X _ _ Z Y Collider Non-Collider X Y Z X Y Z
17 PAG Search: Orientation X1 After Adjacency Phase X3 X4 X2 Collider Test: X1 X3 X2 X1 X1 _ _ X2 X3 X4 X2 Away from Collider Test: X1 X3 X4 X2 X3 X4 X 1 X1 _ _ X4 X3 X2 _ _ X4 X3 X 2 X 3 X 4
18 Interesting Cases M1 L L X1 X2 M2 X Y Z Y1 Y2 L1 Z1 L X Y Z1 X L1 Z2 L2 Z2 M3 Y M4 18
19 Tetrad Demo and Hands-on 1) Create new session 2) Select Search from Simulated Data from Template menu 3) Build graphs for M1, M2, M3 interesting cases, parameterize, instantiate, and generate sample data N=1,000. 4) Execute PC search, a =.05 5) Execute FCI search, a =.05 M1 L X Y Z L M2 X1 Y1 X2 Y2 Z1 L X M3 Y L1 Z2 19
20 Regression & Causal Inference 20
21 Regression & Causal Inference Typical (non-experimental) strategy: 1. Establish a prima facie case (X associated with Y) Z But, omitted variable bias X Y 2. So, identifiy and measure potential confounders Z: a) prior to X, b) associated with X, c) associated with Y 3. Statistically adjust for Z (multiple regression) 21
22 Regression & Causal Inference Multiple regression or any similar strategy is provably unreliable for causal inference regarding X Y, with covariates Z, unless: X truly prior to Y X, Z, and Y are causally sufficient (no confounding) 22
23 Tetrad Demo and Hands-on 1) Create new session 2) Select Search from Simulated Data from Template menu 3) Build a graph for M4 interesting cases, parameterize as SEM, instantiate, and generate sample data N=1,000. 4) Execute PC search, a =.05 5) Execute FCI search, a =.05 23
24 Measurement 24
25 Measurement Error and Coarsening Endanger conditional Independence! X Z Y X _ _ Y Z Z Z* Measurement Error: Z = Z + e X _ _ Y Z (unless Var(e ) = 0) e Coarsening: - < Z < 0 Z* = 0 0 Z < i Z* = 1 i Z < j Z* = 2.. k Z < Z* = k X _ _ Y Z* (almost always) 25
26 Strategies 1. Parameterize measurement error: Sensitivity Analysis Bayesian Analysis Bounds X Z Y Z e 2. Multiple Indicators: X Z Y Z1 Z2 26
27 Strategies 1. Parameterize measurement error: Sensitivity Analysis Bayesian Analysis Bounds X Z Y Z e 2. Multiple Indicators: Scales X Z Y Z1 Z2 X _ _ Y Z_scale Z_scale 27
28 Psuedorandom sample: N = 2,000 Parental Resources Lead Exposure IQ Regression of IQ on Lead, PR Independent Variable Coefficient Estimate p-value PR Lead
29 Multiple Measures of the Confounder e 1 e 2 e 3 X 1 X 2 X 3 Lead Exposure Parental Resources IQ X 1 := g 1 * Parental Resources + e 1 X 2 := g 2 * Parental Resources + e 2 X 3 := g 3 * Parental Resources + e 3 29
30 Scales don't preserve conditional independence X 1 X 2 X 3 Lead Exposure Parental Resources IQ PR_Scale = (X 1 + X 2 + X 3 ) / 3 Independent Variable Coefficient Estimate p-value PR_scale Lead
31 Indicators Don t Preserve Conditional Independence X 1 X 2 X 3 Lead Exposure Parental Resources IQ Regress IQ on: Lead, X 1, X 2, X 3 Independent Variable Coefficient Estimate p-value X X X Lead
32 Strategies 1. Parameterize measurement error: Sensitivity Analysis Bayesian Analysis Bounds 2. Multiple Indicators: Scales SEM ex X Z Y Z e xy X Z Y Z1 Z2 ey ez1 ez2 E( ˆ ) yx 0 X _ _ Y Z 32
33 Structural Equation Models Work True Model Estimated Model X 1 X 2 X 3 X 1 X 2 X 3 Lead Exposure Parental Resources IQ Lead Exposure Parental Resources IQ In the Structural Equation Model E( ˆ) 0 ˆ.07 (p-value =.499) Lead and IQ screened off by PR 33
34 Coarsening is Bad PR_binary Lead Exposure Parental Resources IQ Parental Resources < m(pr) PR_binary = 0 Parental Resources m(pr) PR_binary = 1 Independent Variable Coefficient Estimate p-value Screened-off at.05? PR_binary No Lead No 34
35 Coarsening is Bad PR_binary Lead Exposure Parental Resources IQ Parental Resources < m(pr) PR_binary = 0 Parental Resources m(pr) PR_binary = 1 Independent Variable Coefficient Estimate p-value Screened-off at.05? PR_binary No Lead No 35
36 TV Obesity Proctor, et al. (2003). Television viewing and change in body fat from preschool to early adolescence: The Framingham Children s Study International Journal of Obesity, 27, Exercise TV Obesity (BMI) Diet Goals: Estimate the influence of TV on BMI Tease apart the mechanisms (diet, exercise) 36
37 Measures of Exercise, Diet Exercise_M [L,H] TV (age 4) Exercise Diet (Calories) Obesity (BMI) Age 11 Diet_M [L,H] Exercise_M: L Calories expended in exercise in bottom two tertiles Exercise_M: H Calories expended in exercise in top tertile Diet_M: L Calories consumed in bottom two tertiles Diet_M: H Calories consumed in top tertile 37
38 Measures of Exercise, Diet Exercise_M [L,H] TV (age 4) Exercise Diet (Calories) Obesity (BMI) Age 11 Diet_M [L,H] Findings: TV and Obesity NOT screened off by Exercise_M & Diet_M Bias in mechanism estimation unknown 38
39 Problems with Latent Variable SEMs Specified Model True Model ex xy ey ex ey X Z Y X Z Y Z1 Z2 Z1 Z2 ez1 ez2 ez1 ez2 E( ˆ ) yx 0 X _ _ Y Z E( ˆ yx ) xy 39
40 Latent Variable Models Full Model F1 F2 F3 x 1 x 2 x 3 x 4 y 1 y 2 y 3 y 4 z 1 z 2 z 3 z 4 Structural Model F1 F2 F3 Measurement Model F1 F2 F3 x 1 x 2 x 3 x 4 y 1 y 2 y 3 y 4 z 1 z 2 z 3 z 4 40
41 Psychometric Models Social/Personality Psychology St 1 1 Dep 1 1 St St 21 1 Stress? + Depression? Coping Via Religion Dep Dep 20 1 C 1 C 2.. C 20 41
42 Psychometric Models Educational Research 42
43 Local Independence / Pure Measurement Models Not Locally Independent F1 F2 F3 x 1 x 2 x 3 x 4 y 1 y 2 y 3 y 4 z 1 z 2 z 3 z 4 F Locally Independent F1 F2 F3 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 Local Independence: For every pair of measured items x i, x j : x i _ _ x j modeled latent parents of x i 43
44 Local Independence / Pure Measurement Models Impure F1 F2 F3 x 1 x 2 x 3 x 4 y 1 y 2 y 3 y 4 z 1 z 2 z 3 z 4 F 1 st Order Pure F1 F2 F3 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 44
45 Local Independence / Pure Measurement Models 1 st Order Pure F1 F2 F3 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 2 nd Order Pure F1 F2 F3 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 G 45
46 Rank 1 Constraints: Tetrad Equations Fact: given W L it follows that X Y Z W = 1 L + e 1 X = 2 L + e 2 Y = 3 L + e 3 Z = 4 L + e 4 Cov WX Cov YZ = ( L) ( L) = = ( L) ( L) =Cov WY Cov XZ WX YZ = WY XZ = WZ XY tetrad constraints
47 Charles Spearman (1904) Statistical Constraints Measurement Model Structure r m1,m2 * r r1,r2 = r m1,r1 * r m2,r2 = r m1,r2 * r m2,r1 g m1 m2 r1 r2
48 Impurities/Deviations from Local Independence defeat tetrad constraints selectively Truth F1 Truth F1 x 1 x 2 x 3 x 4 x 1 x 2 x 3 x 4 F5 r x1,x2 * r x3,x4 = r x1,x3 * r x2,x4 r x1,x2 * r x3,x4 = r x1,x4 * r x2,x3 r x1,x3 * r x2,x4 = r x1,x4 * r x2,x3 r x1,x2 * r x3,x4 = r x1,x3 * r x2,x4 r x1,x2 * r x3,x4 = r x1,x4 * r x2,x3 r x1,x3 * r x2,x4 = r x1,x4 * r x2,x3 48
49 Strategies 1. Cluster and Purify MM first 1. Use rank constraints to find item subsets that form n th order pure clusters 2. Using Pure MM : Search for Structural Model by testing independence relations among latents via SEM estimation 2. Specify Impure Measurement Model 1. Specify Measurement Model for all items 2. Using Specified MM: Search for Structural Model by testing independence relations among latents via SEM estimation 49
50 Purify Impure F1 F2 F3 x 1 x 2 x 3 x 4 y 1 y 2 y 3 y 4 z 1 z 2 z 3 z 4 F 1 st Order Purified F1 F2 F3 x 1 x 2 X3 y 1 y 2 y 3 y 4 z 1 z 3 z 4 X 4 Z2 50
51 Search for Measurement Models BPC, FOFC: Find One Factor Clusters Input: Covariance Matrix of measured items: Output: Subset of items and clusters that are 1st Order Pure FTFC: Find Two Factor Clusters Input: Covariance Matrix of measured items: Output: Subset of items and clusters that are 2 nd Order Pure
52 BPC Case Study: Stress, Depression, and Religion Masters Students (N = 127) 61 - item survey (Likert Scale) Stress: St 1 - St 21 Depression: D 1 - D 20 Religious Coping: C 1 - C 20 St 1 1 St St 21 1 Specified Model Depression Stress + + Coping - Dep 1 1 Dep Dep 20 1 P(c2) = 0.00 C 1 C 2.. C 20 52
53 Case Study: Stress, Depression, and Religion St 3 1 Build Pure Clusters St 4 1 St 16 1 St 18 1 Stress Coping Dep 9 1 Depression Dep 13 1 Dep 19 1 St 20 1 C 9 C 12 C 15 C 14 53
54 Case Study: Stress, Depression, and Religion Assume : Stress causally prior to Depression Find : Stress _ _ Coping Depression St 3 1 St 4 1 St 16 1 St 18 1 Stress + Coping Dep 9 1 Depression Dep Dep 19 1 St 20 1 C 9 C 12 C 15 C 14 P(c2) =
55 2 nd Order Pure F1 F2 F3 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 G S(X) FTFC 2 nd -Order Pure Clusters: {X 1, X 2, X 3, X 4, X 5, X 6 } {X 8, X 9, X 10, X 11, X 12 } 55
56 Summary of Search 56
57 Causal Search from Passive Observation PC, FGS Patterns (Markov equivalence class - no latent confounding) FCI PAGs (Markov equivalence - including confounders and selection bias) CCD Linear cyclic models (no confounding) Lingam unique DAG (no confounding linear non-gaussian faithfulness not needed) BPC, FOFC, FTFC (Equivalence class of linear latent variable models) LVLingam set of DAGs (confounders allowed) CyclicLingam set of DGs (cyclic models, no confounding) Non-linear additive noise models unique DAG Most of these algorithms are pointwise consistent uniform consistent algorithms require stronger assumptions 57
58 Causal Search from Manipulations/Interventions What sorts of manipulation/interventions have been studied? Do(X=x) : replace P(X parents(x)) with P(X=x) = 1.0 Randomize(X): (replace P(X parents(x)) with P M (X), e.g., uniform) Soft interventions (replace P(X parents(x)) with P M (X parents(x), I), P M (I)) Simultaneous interventions (reduces the number of experiments required to be guaranteed to find the truth with an independence oracle from N-1 to 2 log(n) Sequential interventions Sequential, conditional interventions Time sensitive interventions 58
Causal Discovery with Latent Variables: the Measurement Problem
Causal Discovery with Latent Variables: the Measurement Problem Richard Scheines Carnegie Mellon University Peter Spirtes, Clark Glymour, Ricardo Silva, Steve Riese, Erich Kummerfeld, Renjie Yang, Max
More informationTutorial: Causal Model Search
Tutorial: Causal Model Search Richard Scheines Carnegie Mellon University Peter Spirtes, Clark Glymour, Joe Ramsey, others 1 Goals 1) Convey rudiments of graphical causal models 2) Basic working knowledge
More informationAutomatic Causal Discovery
Automatic Causal Discovery Richard Scheines Peter Spirtes, Clark Glymour Dept. of Philosophy & CALD Carnegie Mellon 1 Outline 1. Motivation 2. Representation 3. Discovery 4. Using Regression for Causal
More informationCausal Discovery. Richard Scheines. Peter Spirtes, Clark Glymour, and many others. Dept. of Philosophy & CALD Carnegie Mellon
Causal Discovery Richard Scheines Peter Spirtes, Clark Glymour, and many others Dept. of Philosophy & CALD Carnegie Mellon Graphical Models --11/30/05 1 Outline 1. Motivation 2. Representation 3. Connecting
More informationCausal Inference & Reasoning with Causal Bayesian Networks
Causal Inference & Reasoning with Causal Bayesian Networks Neyman-Rubin Framework Potential Outcome Framework: for each unit k and each treatment i, there is a potential outcome on an attribute U, U ik,
More informationMeasurement Error and Causal Discovery
Measurement Error and Causal Discovery Richard Scheines & Joseph Ramsey Department of Philosophy Carnegie Mellon University Pittsburgh, PA 15217, USA 1 Introduction Algorithms for causal discovery emerged
More informationGeneralized Measurement Models
Generalized Measurement Models Ricardo Silva and Richard Scheines Center for Automated Learning and Discovery rbas@cs.cmu.edu, scheines@andrew.cmu.edu May 16, 2005 CMU-CALD-04-101 School of Computer Science
More informationLearning the Structure of Linear Latent Variable Models
Journal (2005) - Submitted /; Published / Learning the Structure of Linear Latent Variable Models Ricardo Silva Center for Automated Learning and Discovery School of Computer Science rbas@cs.cmu.edu Richard
More informationAbstract. Three Methods and Their Limitations. N-1 Experiments Suffice to Determine the Causal Relations Among N Variables
N-1 Experiments Suffice to Determine the Causal Relations Among N Variables Frederick Eberhardt Clark Glymour 1 Richard Scheines Carnegie Mellon University Abstract By combining experimental interventions
More informationCausality in Econometrics (3)
Graphical Causal Models References Causality in Econometrics (3) Alessio Moneta Max Planck Institute of Economics Jena moneta@econ.mpg.de 26 April 2011 GSBC Lecture Friedrich-Schiller-Universität Jena
More informationCausal Reasoning with Ancestral Graphs
Journal of Machine Learning Research 9 (2008) 1437-1474 Submitted 6/07; Revised 2/08; Published 7/08 Causal Reasoning with Ancestral Graphs Jiji Zhang Division of the Humanities and Social Sciences California
More informationChallenges (& Some Solutions) and Making Connections
Challenges (& Some Solutions) and Making Connections Real-life Search All search algorithm theorems have form: If the world behaves like, then (probability 1) the algorithm will recover the true structure
More informationCenter for Causal Discovery: Summer Workshop
Center for Causal Discovery: Summer Workshop - 2015 June 8-11, 2015 Carnegie Mellon University 1 Goals 1) Working knowledge of graphical causal models 2) Basic working knowledge of Tetrad V 3) Basic understanding
More informationarxiv: v4 [math.st] 19 Jun 2018
Complete Graphical Characterization and Construction of Adjustment Sets in Markov Equivalence Classes of Ancestral Graphs Emilija Perković, Johannes Textor, Markus Kalisch and Marloes H. Maathuis arxiv:1606.06903v4
More informationCarnegie Mellon Pittsburgh, Pennsylvania 15213
A Tutorial On Causal Inference Peter Spirtes August 4, 2009 Technical Report No. CMU-PHIL-183 Philosophy Methodology Logic Carnegie Mellon Pittsburgh, Pennsylvania 15213 1. Introduction A Tutorial On Causal
More informationArrowhead completeness from minimal conditional independencies
Arrowhead completeness from minimal conditional independencies Tom Claassen, Tom Heskes Radboud University Nijmegen The Netherlands {tomc,tomh}@cs.ru.nl Abstract We present two inference rules, based on
More informationPDF hosted at the Radboud Repository of the Radboud University Nijmegen
PDF hosted at the Radboud Repository of the Radboud University Nijmegen The following full text is an author's version which may differ from the publisher's version. For additional information about this
More informationEquivalence in Non-Recursive Structural Equation Models
Equivalence in Non-Recursive Structural Equation Models Thomas Richardson 1 Philosophy Department, Carnegie-Mellon University Pittsburgh, P 15213, US thomas.richardson@andrew.cmu.edu Introduction In the
More informationCausal Inference in the Presence of Latent Variables and Selection Bias
499 Causal Inference in the Presence of Latent Variables and Selection Bias Peter Spirtes, Christopher Meek, and Thomas Richardson Department of Philosophy Carnegie Mellon University Pittsburgh, P 15213
More informationSupplementary material to Structure Learning of Linear Gaussian Structural Equation Models with Weak Edges
Supplementary material to Structure Learning of Linear Gaussian Structural Equation Models with Weak Edges 1 PRELIMINARIES Two vertices X i and X j are adjacent if there is an edge between them. A path
More informationCausal Structure Learning and Inference: A Selective Review
Vol. 11, No. 1, pp. 3-21, 2014 ICAQM 2014 Causal Structure Learning and Inference: A Selective Review Markus Kalisch * and Peter Bühlmann Seminar for Statistics, ETH Zürich, CH-8092 Zürich, Switzerland
More informationLearning causal network structure from multiple (in)dependence models
Learning causal network structure from multiple (in)dependence models Tom Claassen Radboud University, Nijmegen tomc@cs.ru.nl Abstract Tom Heskes Radboud University, Nijmegen tomh@cs.ru.nl We tackle the
More informationANALYTIC COMPARISON. Pearl and Rubin CAUSAL FRAMEWORKS
ANALYTIC COMPARISON of Pearl and Rubin CAUSAL FRAMEWORKS Content Page Part I. General Considerations Chapter 1. What is the question? 16 Introduction 16 1. Randomization 17 1.1 An Example of Randomization
More informationIntroduction to Causal Calculus
Introduction to Causal Calculus Sanna Tyrväinen University of British Columbia August 1, 2017 1 / 1 2 / 1 Bayesian network Bayesian networks are Directed Acyclic Graphs (DAGs) whose nodes represent random
More informationUsing Descendants as Instrumental Variables for the Identification of Direct Causal Effects in Linear SEMs
Using Descendants as Instrumental Variables for the Identification of Direct Causal Effects in Linear SEMs Hei Chan Institute of Statistical Mathematics, 10-3 Midori-cho, Tachikawa, Tokyo 190-8562, Japan
More informationInterpreting and using CPDAGs with background knowledge
Interpreting and using CPDAGs with background knowledge Emilija Perković Seminar for Statistics ETH Zurich, Switzerland perkovic@stat.math.ethz.ch Markus Kalisch Seminar for Statistics ETH Zurich, Switzerland
More informationCOMP538: Introduction to Bayesian Networks
COMP538: Introduction to Bayesian Networks Lecture 9: Optimal Structure Learning Nevin L. Zhang lzhang@cse.ust.hk Department of Computer Science and Engineering Hong Kong University of Science and Technology
More informationLearning equivalence classes of acyclic models with latent and selection variables from multiple datasets with overlapping variables
Learning equivalence classes of acyclic models with latent and selection variables from multiple datasets with overlapping variables Robert E. Tillman Carnegie Mellon University Pittsburgh, PA rtillman@cmu.edu
More informationPairwise Cluster Comparison for Learning Latent Variable Models
Pairwise Cluster Comparison for Learning Latent Variable Models Nuaman Asbeh Industrial Engineering and Management Ben-Gurion University of the Negev Beer-Sheva, 84105, Israel Boaz Lerner Industrial Engineering
More informationAcyclic Linear SEMs Obey the Nested Markov Property
Acyclic Linear SEMs Obey the Nested Markov Property Ilya Shpitser Department of Computer Science Johns Hopkins University ilyas@cs.jhu.edu Robin J. Evans Department of Statistics University of Oxford evans@stats.ox.ac.uk
More informationarxiv: v3 [stat.me] 3 Jun 2015
The Annals of Statistics 2015, Vol. 43, No. 3, 1060 1088 DOI: 10.1214/14-AOS1295 c Institute of Mathematical Statistics, 2015 A GENERALIZED BACK-DOOR CRITERION 1 arxiv:1307.5636v3 [stat.me] 3 Jun 2015
More informationInterventions and Causal Inference
Interventions and Causal Inference Frederick Eberhardt 1 and Richard Scheines Department of Philosophy Carnegie Mellon University Abstract The literature on causal discovery has focused on interventions
More informationGraphical models and causality: Directed acyclic graphs (DAGs) and conditional (in)dependence
Graphical models and causality: Directed acyclic graphs (DAGs) and conditional (in)dependence General overview Introduction Directed acyclic graphs (DAGs) and conditional independence DAGs and causal effects
More informationThe TETRAD Project: Constraint Based Aids to Causal Model Specification
The TETRAD Project: Constraint Based Aids to Causal Model Specification Richard Scheines, Peter Spirtes, Clark Glymour, Christopher Meek and Thomas Richardson Department of Philosophy, Carnegie Mellon
More informationUsing background knowledge for the estimation of total causal e ects
Using background knowledge for the estimation of total causal e ects Interpreting and using CPDAGs with background knowledge Emilija Perkovi, ETH Zurich Joint work with Markus Kalisch and Marloes Maathuis
More informationCausal Clustering for 1-Factor Measurement Models
Causal Clustering for -Factor Measurement Models Erich Kummerfeld University of Pittsburgh 5607 Baum Boulevard, Suite 500 Pittsburgh, PA 5206 erk57@pitt.edu Joseph Ramsey Carnegie Mellon University 39
More informationCausal Inference for High-Dimensional Data. Atlantic Causal Conference
Causal Inference for High-Dimensional Data Atlantic Causal Conference Overview Conditional Independence Directed Acyclic Graph (DAG) Models factorization and d-separation Markov equivalence Structure Learning
More informationControlling for latent confounding by confirmatory factor analysis (CFA) Blinded Blinded
Controlling for latent confounding by confirmatory factor analysis (CFA) Blinded Blinded 1 Background Latent confounder is common in social and behavioral science in which most of cases the selection mechanism
More informationLearning Measurement Models for Unobserved Variables
UAI2003 SILVA ET AL. 543 Learning Measurement Models for Unobserved Variables Ricardo Silva, Richard Scheines, Clark Glymour and Peter Spirtes* Center for Automated Learning and Discovery Carnegie Mellon
More informationLearning Latent Variable Models by Pairwise Cluster Comparison: Part I Theory and Overview
Journal of Machine Learning Research 17 2016 1-52 Submitted 9/14; Revised 5/15; Published 12/16 Learning Latent Variable Models by Pairwise Cluster Comparison: Part I Theory and Overview Nuaman Asbeh Department
More informationLEARNING CAUSAL EFFECTS: BRIDGING INSTRUMENTS AND BACKDOORS
LEARNING CAUSAL EFFECTS: BRIDGING INSTRUMENTS AND BACKDOORS Ricardo Silva Department of Statistical Science and Centre for Computational Statistics and Machine Learning, UCL Fellow, Alan Turing Institute
More informationIdentifiability assumptions for directed graphical models with feedback
Biometrika, pp. 1 26 C 212 Biometrika Trust Printed in Great Britain Identifiability assumptions for directed graphical models with feedback BY GUNWOONG PARK Department of Statistics, University of Wisconsin-Madison,
More informationUndirected Graphical Models
Undirected Graphical Models 1 Conditional Independence Graphs Let G = (V, E) be an undirected graph with vertex set V and edge set E, and let A, B, and C be subsets of vertices. We say that C separates
More informationCAUSAL DISCOVERY UNDER NON-STATIONARY FEEDBACK
CAUSAL DISCOVERY UNDER NON-STATIONARY FEEDBACK by Eric V. Strobl B.A. in Molecular & Cell Biology, University of California at Berkeley, 2009 B.A. in Psychology, University of California at Berkeley, 2009
More informationChris Bishop s PRML Ch. 8: Graphical Models
Chris Bishop s PRML Ch. 8: Graphical Models January 24, 2008 Introduction Visualize the structure of a probabilistic model Design and motivate new models Insights into the model s properties, in particular
More informationCausal Effect Identification in Alternative Acyclic Directed Mixed Graphs
Proceedings of Machine Learning Research vol 73:21-32, 2017 AMBN 2017 Causal Effect Identification in Alternative Acyclic Directed Mixed Graphs Jose M. Peña Linköping University Linköping (Sweden) jose.m.pena@liu.se
More informationProbabilistic Causal Models
Probabilistic Causal Models A Short Introduction Robin J. Evans www.stat.washington.edu/ rje42 ACMS Seminar, University of Washington 24th February 2011 1/26 Acknowledgements This work is joint with Thomas
More informationLearning Semi-Markovian Causal Models using Experiments
Learning Semi-Markovian Causal Models using Experiments Stijn Meganck 1, Sam Maes 2, Philippe Leray 2 and Bernard Manderick 1 1 CoMo Vrije Universiteit Brussel Brussels, Belgium 2 LITIS INSA Rouen St.
More informationJOINT PROBABILISTIC INFERENCE OF CAUSAL STRUCTURE
JOINT PROBABILISTIC INFERENCE OF CAUSAL STRUCTURE Dhanya Sridhar Lise Getoor U.C. Santa Cruz KDD Workshop on Causal Discovery August 14 th, 2016 1 Outline Motivation Problem Formulation Our Approach Preliminary
More informationGraphical Representation of Causal Effects. November 10, 2016
Graphical Representation of Causal Effects November 10, 2016 Lord s Paradox: Observed Data Units: Students; Covariates: Sex, September Weight; Potential Outcomes: June Weight under Treatment and Control;
More informationarxiv: v2 [cs.lg] 9 Mar 2017
Journal of Machine Learning Research? (????)??-?? Submitted?/??; Published?/?? Joint Causal Inference from Observational and Experimental Datasets arxiv:1611.10351v2 [cs.lg] 9 Mar 2017 Sara Magliacane
More informationCS Lecture 3. More Bayesian Networks
CS 6347 Lecture 3 More Bayesian Networks Recap Last time: Complexity challenges Representing distributions Computing probabilities/doing inference Introduction to Bayesian networks Today: D-separation,
More informationPDF hosted at the Radboud Repository of the Radboud University Nijmegen
PDF hosted at the Radboud Repository of the Radboud University Nijmegen The following full text is a preprint version which may differ from the publisher's version. For additional information about this
More informationLearning in Bayesian Networks
Learning in Bayesian Networks Florian Markowetz Max-Planck-Institute for Molecular Genetics Computational Molecular Biology Berlin Berlin: 20.06.2002 1 Overview 1. Bayesian Networks Stochastic Networks
More informationCausal Bayesian networks. Peter Antal
Causal Bayesian networks Peter Antal antal@mit.bme.hu A.I. 11/25/2015 1 Can we represent exactly (in)dependencies by a BN? From a causal model? Suff.&nec.? Can we interpret edges as causal relations with
More informationCausation and Intervention. Frederick Eberhardt
Causation and Intervention Frederick Eberhardt Abstract Accounts of causal discovery have traditionally split into approaches based on passive observational data and approaches based on experimental interventions
More informationBayesian Networks (Part II)
10-601 Introduction to Machine Learning Machine Learning Department School of Computer Science Carnegie Mellon University Bayesian Networks (Part II) Graphical Model Readings: Murphy 10 10.2.1 Bishop 8.1,
More informationRecall, Positive/Negative Association:
ANNOUNCEMENTS: Remember that discussion today is not for credit. Go over R Commander. Go to 192 ICS, except at 4pm, go to 192 or 174 ICS. TODAY: Sections 5.3 to 5.5. Note this is a change made in the daily
More informationCausal Discovery by Computer
Causal Discovery by Computer Clark Glymour Carnegie Mellon University 1 Outline 1. A century of mistakes about causation and discovery: 1. Fisher 2. Yule 3. Spearman/Thurstone 2. Search for causes is statistical
More informationCausality. Pedro A. Ortega. 18th February Computational & Biological Learning Lab University of Cambridge
Causality Pedro A. Ortega Computational & Biological Learning Lab University of Cambridge 18th February 2010 Why is causality important? The future of machine learning is to control (the world). Examples
More informationCausal Discovery in the Presence of Measurement Error: Identifiability Conditions
Causal Discovery in the Presence of Measurement Error: Identifiability Conditions Kun Zhang,MingmingGong?,JosephRamsey,KayhanBatmanghelich?, Peter Spirtes, Clark Glymour Department of philosophy, Carnegie
More informationFaithfulness of Probability Distributions and Graphs
Journal of Machine Learning Research 18 (2017) 1-29 Submitted 5/17; Revised 11/17; Published 12/17 Faithfulness of Probability Distributions and Graphs Kayvan Sadeghi Statistical Laboratory University
More informationCausal Inference on Data Containing Deterministic Relations.
Causal Inference on Data Containing Deterministic Relations. Jan Lemeire Kris Steenhaut Sam Maes COMO lab, ETRO Department Vrije Universiteit Brussel, Belgium {jan.lemeire, kris.steenhaut}@vub.ac.be, sammaes@gmail.com
More informationarxiv: v3 [stat.me] 10 Mar 2016
Submitted to the Annals of Statistics ESTIMATING THE EFFECT OF JOINT INTERVENTIONS FROM OBSERVATIONAL DATA IN SPARSE HIGH-DIMENSIONAL SETTINGS arxiv:1407.2451v3 [stat.me] 10 Mar 2016 By Preetam Nandy,,
More informationLearning Instrumental Variables with Structural and Non-Gaussianity Assumptions
Journal of Machine Learning Research 18 (2017) 1-49 Submitted 1/17; Revised 8/17; Published 11/17 Learning Instrumental Variables with Structural and Non-Gaussianity Assumptions Ricardo Silva ricardo@stats.ucl.ac.u
More informationIntroduction to Probabilistic Graphical Models
Introduction to Probabilistic Graphical Models Sargur Srihari srihari@cedar.buffalo.edu 1 Topics 1. What are probabilistic graphical models (PGMs) 2. Use of PGMs Engineering and AI 3. Directionality in
More informationCausal Models with Hidden Variables
Causal Models with Hidden Variables Robin J. Evans www.stats.ox.ac.uk/ evans Department of Statistics, University of Oxford Quantum Networks, Oxford August 2017 1 / 44 Correlation does not imply causation
More informationProbabilistic Graphical Models
Probabilistic Graphical Models Brown University CSCI 295-P, Spring 213 Prof. Erik Sudderth Lecture 11: Inference & Learning Overview, Gaussian Graphical Models Some figures courtesy Michael Jordan s draft
More informationSingle World Intervention Graphs (SWIGs):
Single World Intervention Graphs (SWIGs): Unifying the Counterfactual and Graphical Approaches to Causality Thomas Richardson Department of Statistics University of Washington Joint work with James Robins
More informationNoisy-OR Models with Latent Confounding
Noisy-OR Models with Latent Confounding Antti Hyttinen HIIT & Dept. of Computer Science University of Helsinki Finland Frederick Eberhardt Dept. of Philosophy Washington University in St. Louis Missouri,
More informationNew developments in structural equation modeling
New developments in structural equation modeling Rex B Kline Concordia University Montréal Set A: SCM A1 UNL Methodology Workshop A2 A3 A4 Topics o Graph theory o Mediation: Design Conditional Causal A5
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning Mark Schmidt University of British Columbia Winter 2018 Last Time: Learning and Inference in DAGs We discussed learning in DAG models, log p(x W ) = n d log p(x i j x i pa(j),
More informationCausal Bayesian networks. Peter Antal
Causal Bayesian networks Peter Antal antal@mit.bme.hu A.I. 4/8/2015 1 Can we represent exactly (in)dependencies by a BN? From a causal model? Suff.&nec.? Can we interpret edges as causal relations with
More informationPDF hosted at the Radboud Repository of the Radboud University Nijmegen
PDF hosted at the Radboud Repository of the Radboud University Nijmegen The following full text is a preprint version which may differ from the publisher's version. For additional information about this
More informationStructural equation modeling
Structural equation modeling Rex B Kline Concordia University Montréal ISTQL Set B B1 Data, path models Data o N o Form o Screening B2 B3 Sample size o N needed: Complexity Estimation method Distributions
More informationBayesian Networks Introduction to Machine Learning. Matt Gormley Lecture 24 April 9, 2018
10-601 Introduction to Machine Learning Machine Learning Department School of Computer Science Carnegie Mellon University Bayesian Networks Matt Gormley Lecture 24 April 9, 2018 1 Homework 7: HMMs Reminders
More informationIdentifying Linear Causal Effects
Identifying Linear Causal Effects Jin Tian Department of Computer Science Iowa State University Ames, IA 50011 jtian@cs.iastate.edu Abstract This paper concerns the assessment of linear cause-effect relationships
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning Undirected Graphical Models Mark Schmidt University of British Columbia Winter 2016 Admin Assignment 3: 2 late days to hand it in today, Thursday is final day. Assignment 4:
More informationarxiv: v1 [cs.ai] 16 Sep 2015
Causal Model Analysis using Collider v-structure with Negative Percentage Mapping arxiv:1509.090v1 [cs.ai 16 Sep 015 Pramod Kumar Parida Electrical & Electronics Engineering Science University of Johannesburg
More informationData Mining 2018 Bayesian Networks (1)
Data Mining 2018 Bayesian Networks (1) Ad Feelders Universiteit Utrecht Ad Feelders ( Universiteit Utrecht ) Data Mining 1 / 49 Do you like noodles? Do you like noodles? Race Gender Yes No Black Male 10
More informationGeneralized Do-Calculus with Testable Causal Assumptions
eneralized Do-Calculus with Testable Causal Assumptions Jiji Zhang Division of Humanities and Social Sciences California nstitute of Technology Pasadena CA 91125 jiji@hss.caltech.edu Abstract A primary
More informationGaussian Processes. Le Song. Machine Learning II: Advanced Topics CSE 8803ML, Spring 2012
Gaussian Processes Le Song Machine Learning II: Advanced Topics CSE 8803ML, Spring 01 Pictorial view of embedding distribution Transform the entire distribution to expected features Feature space Feature
More informationSEM with observed variables: parameterization and identification. Psychology 588: Covariance structure and factor models
SEM with observed variables: parameterization and identification Psychology 588: Covariance structure and factor models Limitations of SEM as a causal modeling 2 If an SEM model reflects the reality, the
More information1. what conditional independencies are implied by the graph. 2. whether these independecies correspond to the probability distribution
NETWORK ANALYSIS Lourens Waldorp PROBABILITY AND GRAPHS The objective is to obtain a correspondence between the intuitive pictures (graphs) of variables of interest and the probability distributions of
More informationSampling Rejection Sampling Importance Sampling Markov Chain Monte Carlo. Sampling Methods. Oliver Schulte - CMPT 419/726. Bishop PRML Ch.
Sampling Methods Oliver Schulte - CMP 419/726 Bishop PRML Ch. 11 Recall Inference or General Graphs Junction tree algorithm is an exact inference method for arbitrary graphs A particular tree structure
More informationGov 2002: 4. Observational Studies and Confounding
Gov 2002: 4. Observational Studies and Confounding Matthew Blackwell September 10, 2015 Where are we? Where are we going? Last two weeks: randomized experiments. From here on: observational studies. What
More informationGRAPHICAL MARKOV MODELS,
7 th International Conference on Computational and Methodological Statistics (ERCIM 2014), 6-8 December 2014, University of Pisa, Italy: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Three sessions
More informationThe International Journal of Biostatistics
The International Journal of Biostatistics Volume 7, Issue 1 2011 Article 16 A Complete Graphical Criterion for the Adjustment Formula in Mediation Analysis Ilya Shpitser, Harvard University Tyler J. VanderWeele,
More informationProbabilistic Graphical Networks: Definitions and Basic Results
This document gives a cursory overview of Probabilistic Graphical Networks. The material has been gleaned from different sources. I make no claim to original authorship of this material. Bayesian Graphical
More informationDetecting marginal and conditional independencies between events and learning their causal structure.
Detecting marginal and conditional independencies between events and learning their causal structure. Jan Lemeire 1,4, Stijn Meganck 1,3, Albrecht Zimmermann 2, and Thomas Dhollander 3 1 ETRO Department,
More informationProbabilistic Graphical Models (I)
Probabilistic Graphical Models (I) Hongxin Zhang zhx@cad.zju.edu.cn State Key Lab of CAD&CG, ZJU 2015-03-31 Probabilistic Graphical Models Modeling many real-world problems => a large number of random
More information1 h 9 e $ s i n t h e o r y, a p p l i c a t i a n
T : 99 9 \ E \ : \ 4 7 8 \ \ \ \ - \ \ T \ \ \ : \ 99 9 T : 99-9 9 E : 4 7 8 / T V 9 \ E \ \ : 4 \ 7 8 / T \ V \ 9 T - w - - V w w - T w w \ T \ \ \ w \ w \ - \ w \ \ w \ \ \ T \ w \ w \ w \ w \ \ w \
More informationLearning from Sensor Data: Set II. Behnaam Aazhang J.S. Abercombie Professor Electrical and Computer Engineering Rice University
Learning from Sensor Data: Set II Behnaam Aazhang J.S. Abercombie Professor Electrical and Computer Engineering Rice University 1 6. Data Representation The approach for learning from data Probabilistic
More informationHOMEWORK (due Wed, Jan 23): Chapter 3: #42, 48, 74
ANNOUNCEMENTS: Grades available on eee for Week 1 clickers, Quiz and Discussion. If your clicker grade is missing, check next week before contacting me. If any other grades are missing let me know now.
More informationCausal Discovery with Linear Non-Gaussian Models under Measurement Error: Structural Identifiability Results
Causal Discovery with Linear Non-Gaussian Models under Measurement Error: Structural Identifiability Results Kun Zhang,MingmingGong?, Joseph Ramsey,KayhanBatmanghelich?, Peter Spirtes,ClarkGlymour Department
More informationCORRELATION. compiled by Dr Kunal Pathak
CORRELATION compiled by Dr Kunal Pathak Flow of Presentation Definition Types of correlation Method of studying correlation a) Scatter diagram b) Karl Pearson s coefficient of correlation c) Spearman s
More informationStrategies for Discovering Mechanisms of Mind using fmri: 6 NUMBERS. Joseph Ramsey, Ruben Sanchez Romero and Clark Glymour
1 Strategies for Discovering Mechanisms of Mind using fmri: 6 NUMBERS Joseph Ramsey, Ruben Sanchez Romero and Clark Glymour 2 The Numbers 20 50 5024 9205 8388 500 3 fmri and Mechanism From indirect signals
More informationAxioms of Probability? Notation. Bayesian Networks. Bayesian Networks. Today we ll introduce Bayesian Networks.
Bayesian Networks Today we ll introduce Bayesian Networks. This material is covered in chapters 13 and 14. Chapter 13 gives basic background on probability and Chapter 14 talks about Bayesian Networks.
More informationUncertain Reasoning. Environment Description. Configurations. Models. Bayesian Networks
Bayesian Networks A. Objectives 1. Basics on Bayesian probability theory 2. Belief updating using JPD 3. Basics on graphs 4. Bayesian networks 5. Acquisition of Bayesian networks 6. Local computation and
More informationREVIEW 8/2/2017 陈芳华东师大英语系
REVIEW Hypothesis testing starts with a null hypothesis and a null distribution. We compare what we have to the null distribution, if the result is too extreme to belong to the null distribution (p
More information