STAT231 Lecture : chapter8_causality.pdf
17 Jan 2014
School
Department
Course
Professor
Document Summary
Smoking causes lung cancer : the relationship is not deterministic. Not everybody who smokes gets lung cancer. Association: could look at measures of association, for categorical variables. P(lung cancer| smoker) / p(lung cancer/ non- smoker) Association for continuous variables: (linear )correlation, the nature of correlation is given in regression models y=a + bx +r , r~g(0,sigma, we can test the hypothesis that two variables are correlated. Association is not causation: even if the coefficient b was significant, this is only an association, not a causal relationship. How can we explain this association: a baby"s weight increases over time, retail price index increases over time also (inflation) Confounding variate: a variable z is a confounding variable if it is associated with both x and y. Confounding variables in practice: adjusting for confounding variables is often very hard, because you need to. Paired t-test: two observations are blocked: replication, randomization.
