SSR 2024 Lecture 20 Critical Thinking About Causality PDF

“Let them first discuss a more simple question, namely, the operations of body and brute unintelligent matter; and try whether they can there form any idea of causation and necessity, except that...

“Let them first discuss a more simple question, namely, the operations of body and brute unintelligent matter; and try whether they can there form any idea of causation and necessity, except that of a constant conjunction of objects, and subsequent inference of the mind from one to another.” (Hume, 1777) Scientific & Statistical Reasoning Lecture 20: Critical thinking about causality 29-10-2024 1 Deus ex Ha An Argument for Divine Providence, taken from the Constant Regularity observed in the Births of both Sexes. By Dr. John Arbuthnot, Physician in Ordinary to her Majesty, and Fellow of the College of Physicians and the Royal Society. 1 2 Deus ex Ha 3 Lectu Topic Block 3: Scientific and statistical reasoning about psychological Readings WA constructs Tutorial re # 20 Critical thinking about causality Foster (2010) Pearl & MacKenzie (2018), ch 4 WA9: CT Causality Shadish (2007) 21 Correlation & regression Field ch 7 & 8 22 Multiple regression Field ch 6 & 9 WA10: Correlation, T7: Causal pitfalls Regression & CT: 23 Critical thinking about psychological theory Marewski & Olsson (2009) Dennis & Kintsch (2008) Theory Dienes (2008) ch 1 24 Critical thinking about psychological constructs Blanton & Jaccard (2006) Hughes (2017) WA11: Mediation, Lebel & Peters (2011) Moderation & CT: T8: Theory evaluation & Constructs multiple regression 25 Mediation Field ch 10 26 Moderation Field ch 10 27 Critical thinking about psychological measurement Furr & Bacharach (2014) ch 2 Lord (1953) WA12: CT Measurement, 28 Critical thinking about Individual differences & Cronbach (1957) Individual differences & T9: Escape room mechanisms Kievit et al (2013) Qualitative research 29 Critical thinking about Qualitative vs Quantative Coyle (2015) Hughes (2017) research Block 3 Q&A Interim exam 3, roughly 40% stats & 60% about the rest Today 'Correlation does not imply causation' What is a causal connection? Causal reasoning errors What is a counterfactual? Counterfactuals in different designs Threats of causal inference Beyond 'correlation does not imply causation'. Causal Diagram: Specifying Assumptions Causal Diagram: Identifying Confounds Causal Diagram: Applied 5 Confused about causality “Time for a quick reality check. Despite the hysteria from the political class and the media, smoking doesn't kill. In fact, 2 out of every three smokers does not die from a smoking related illness and 9 out of ten smokers do not contract lung cancer. This is not to say that smoking is good for you... news flash: smoking is not good for you. If you are reading this article through the blue haze of cigarette smoke you should quit.” 6 Causality What is the cause? “Let them first discuss a more simple question, namely, the operations of body and brute unintelligent matter; and try whether they can there form any idea of causation and necessity, except that of a constant conjunction of objects, and subsequent inference of the mind from one to another.” (Hume, 1777) 7 Causality What is the cause? (Shadish, 2007; Wegner & Wheatley, 1999) ‘ That which produces any simple or complex idea, we denote by the general name cause, and that which is produced, effect’ (Locke, in Shadish, 2007, p38) Criteria John Stuart Mill: X causes Y if and only if: Priority: Change X precedes change Y Consistency: Change X varies systematically with change Y Exclusivity: There is no alternative explanation for the relationship 8 Causality Mistaking What is a cause? (Stanovich, 2010) correlation for causation! Problematic academic achievements, drug abuse, pregnancy at a very young age related to low self-esteem Criteria: P1 X correlates with Y Priority P2 If X correlates with Consistency Y, X is the cause of Y Exclusivity C X is the cause of Y If we create a stronger positive sense of self-esteem, those other problems will disappear by themselves. 9 Causality What is a cause? (Stanovich, 2010) People with poor reading skills make more erroneous eye movements, go back to the beginning of the sentence more often (regression) and have more fixations per line of text. Criteria: ? Priority Consistency ? Exclusivity Abnormalities in eye movements (oculomotor skills) cause poorer reading skills 10 Causality Post hoc ergo What is a cause? (Stanovich,2010) propter hoc! Criteria: Priority P1 X precedes Y Consistency P2 If X precedes Y, X is the Exclusivity cause of Y C X is the cause of Y 11 Causality Inversion of cause What is a cause? and effect Criteria: P1 X causes Y P2 If X causes Y, then no Y without X Priority C Without X, no Y Consistency Exclusivity 12  Causality What is a cause? (Shadish,2007;Stanovich,2010) "Weapons don't kill people, but people kill people. What's the cause? Is there only one cause? Often highly polarized political debates about what the 'real' cause of something is Are there single causes that are , strictly speaking, both sufficient and necessary? Research is about INUS conditions, not about ultimate causes 13 Causality What is a cause? INUS condition: a match does not lead to fire without Insufficient other conditions (e.g. oxygen). but Non-redundant substantially different from situation without a match part of an Unnecessary Combination of paper, oxygen, but matches, friction is sufficient to produce fire) Sufficient Condition Which is not necessary: other combos are also possible (sunlight, dry grass, oxygen) 14  Causality What is a cause? The question then becomes how we can check for that non- redundancy In research we try to compare observations we make with a good counterfactual: A perfect counterfactual is knowledge of what would have happened to each participant if they had not undergone a certain manipulation. If we compare that knowledge with what actually happened, we know what the effect of the manipulation is. That "perfect" variant is physically impossible. So what can we do instead? 15 Causality Counterfactuals Research into the relative contribution of CDR vs. affiliation in initiations E.g. Existing groups: student association with heavy hazing, with mild hazing, without hazing What is the counterfactual? Possible threats? 16 Causality Threats (Shadish, 2007; Stanovich, 2010) Ruling out alternative explanations, so whether effect of treatment is confused with…: Outside factors: History: Influences (outside of intervention) over the course of the research, which influence outcome Maturation: Natural changes that may be confused with effect treatment Effects of selection: Selection: Selection criteria for treatment related to outcomes of treatment Attrition: Participants’ dropout, systematically correlated with conditions Unintended effects of study itself: Instrumentation: Change in measuring instrument resulting in a difference between pre- and post-measurement Testing: Effect of measurement itself on subsequent measurements (fatigue, habituation etc) Statistical artifacts Regression to the mean: Extreme scores will be followed by less extreme scores 17  Causality Counterfactuals E.g. Research on the role of Cognitive Dissonance Reductionin hazing/initiations E.g. 3 conditions (between subjects): mild, intense or no hazing/initiation What is the counterfactual? 18 Causal Diagram (Pearl, 2009/2018) Would Family at (not) like the corps to be alone Initiation Fraternizati intensity on 19 Causal Diagram (Pearl, 2009/2018) Would Family Family at the (not) Would (not) atcorps the like to be like to alone corps be alone Initiation Initiation Fraterniza intensity intensity tion 20 Today 'Correlation does not imply causation' What is a causal connection? Causal reasoning errors What is a counterfactual? Counterfactuals in different designs Threats of causal inference Beyond 'correlation does not imply causation'. Causal Diagram: Specifying Assumptions Causal Diagram: Identifying Confounds Causal Diagram: Applied 21  Swipe left? I think it's important that the people I'm dating are attractive or nice. Of the people I select from Tinder for a date it seems like they get more annoying the more attractive they are. And the nice persons are often unattractive. So I think a beautiful personality and a beautiful face are mutually exclusive. 22 Causality correlation & causation Foster (2010) According to Foster (2010), a 'swamp of ambiguity' has arisen (in developmental psychology) around statements about causality. (1) Ignoring Causality Some authors write down only the correlations they find, without making any statements about causality. (2) Statements of causality, but unclear assumptions Other researchers make statements about causal relationships based on correlational data, but often without specifying assumptions. Conceptual: which third variables should be included in the model and why not? (3) Pseudo-correlational statements No direct statements about causality, but clearly the implied in the conclusion Highlight from Y1 Psych: 'The role of attachment style on relationship satisfaction'. 23 Causal Diagram X Y X Y 24 Causal Diagram Z Z X Y X Y X and Y have Effect of X on Y is common cause mediated by Z Z X Y X and Y have a common effect 25 Closing the backdoor Z X Y 26 Causality correlation & causation Meehl (1971); Spetor & Brannick (2010) Purification principle: The mistaken idea that the AZ B R more control variables are included in a model, X Y W the more accurate the estimation of the causal effect is. Problem of overcorrection: controlling for mediators on the causal path could lead to an underestimation of the total causal effect Collider bias: controlling for common effects will bias the estimation of a causal relationship between two variables If, in principle, all confounders are controlled for, a correlation between treatment and outcome can X Y be seen as causal. 27 Purification? Meehl (1971) Class (or SES) Social Schizoph activity renia 28 Purification? Meehl (1971) 1. Lower-class students have less money to spend, hence join fewer activity groups, hence suffer more social isolation, which isolation helps precipitate schizophrenia. 2. Lower-class students are perceived by peers as lower-class, hence snobbishly rejected from activity groups on a class basis, hence suffer more social isolation, which... Social activity 3. Lower-class students tend to acquire less competent social skills in home and neighborhood, hence tend to be peer-rejected (but not on an explicit class basis), hence... Class (or Schizophr SES) enia 29  Gender Gender Length Length of Skull Skull width ofskull skull width 30 Common Cause Conditioning on Income common cause Number of Use of household contraceptives appliances Spurious/biased X and Y are caused by Z 31  Age Background Self- Slang disclosure 32 Mediator people dancing a Conditioning on mediator lot/little Music Temperature good/bad on dance floor Effect of X on Y is mediated by Z 33 Game 1 Testost Baldness Headgear erone Insecurity 34 Collider COVID19 test Conditioning on collider X and Y cause Z Smoking COVID19 Biased 35 Collider (Glymour, 2006) Z: Wet Conditioning on lawn collider/shared consequence X: Y: RAIN SPRINKLER Spurieus/biased X and Y cause Z 36 Cause 1 v Cause 2 37 Cause 1 v Cause 2 38 Collider (Glymour, 2006) NBA basketball Length Speed X and Y cause Z 39 Length v Speed 40 Collider (Glymour, 2006) Hollywood star Acting skills Attractiveness X and Y cause Z 41 Skills v Looks 42 Collider (Glymour, 2006) Restaurant visit X and Y cause Z Hamburger Fries 43 Burger v Fries 44  X Y Z P (X, Y, Z) Heads Heads 1.25 Heads Tails 1.25 Tails Heads 1.25 Tails Tails 0.25 X Y P(X,Y|Z=1) X Y P(X,Y|Z=0) Heads Heads.333 Tails Tails 1 Heads Tails.333 X and Y are Tails Heads.333 dependent if we condition on Z (Pearl, Glymour & Jewell, 2016) 45 46 ` https://watzilei.com/shiny/collider/ 47 Causal Diagram Confound vs Collider (Spector & Brannick, 2010) Liking employees Liking by manager employees by manager Performance Performance Beer at evaluation by Beer at evaluation by office party manager office party manager 48  Confounder vs Collider (Meehl, 1969, Spector & Brannick, 2010) So..: Whether or not to condition on a third variable depends on how you think these variables are actually related to each other This assumption is crucial for how you further analyse the data, and needs to be substantiated. Based on prior research Based on common assumptions The fact that you condition on a third variable must therefore be explicitly stated and substantiated. A causal diagram s an insightful way to make these assumptions explicit. 49 Collider Bias I think it's important that the people I'm dating are attractive or nice. Of the people I select from Tinder for a date it seems like they get more annoying the more attractive they are. And the nice persons are often unattractive. So I think a beautiful personality and a beautiful face are mutually exclusive. 50 Collider (Glymour, 2006) Tinder date Conditioning on collider/common effect Attractiveness Personality Spurieus/biased X and Y cause Z 51 Personality v Looks 52 So You know when we speak of a cause-effect relationship (Mill criteria) You know what an INUS condition is and can apply it to reasoning about causes of psychological states. You know what a counterfactual is & where it is derived from in different research designs You know the most important threats of causal inference in Quasi exp research and can apply them. You understand the three main ways in which a third variable can give a distorted picture of the relationship between two variables Difference between a mediator, common cause (confound), common effect (collider) You can recognize them in a Causal Diagram 53 Bonus Berksons paradox (Pearl & Mackenzie 2018) In hospital Conditioning on collider/common effect Disease A Disease B X and Y cause Z 54 Bonus Mediator & Collider (Pearl & Mackenzie 2018) Birth defect Birth weight Child mortality Smoking 55 Bonus Monty Hall (Pearl & Mackenzie 2018) Door that Conditioning on host opens collider/common effect Door that Door behind which participant is the car chooses X and Y cause Z 56 Bonus Berkeley admissions paradox (Pearl & Mackenzie 2018) Department Conditioning on mediator Gender Admission outcome 57