PO218 Lecture Notes - Lecture 14: Partial Correlation, Spurious Relationship, Causal Model
Document Summary
Chapter 14: partial correlation and multiple regression and correlation. Partial correlation: a multivariate technique for examining a bivariate relationship while controlling for other variables. Control variables: a third variable (z) that might affect a bivariate relationship. It is done by computing pearson"s r for the bivariate relationship and then computing the partial (or first order) correlation coefficient. The independent and dependent are not casually related: intervening relationship: a multivariate relationship in which the independent and dependent variables are linked primarily through control variable, formula for partial relationships, ryx. z = (cid:3045)(cid:3300)(cid:3299) (cid:4666)(cid:3045)(cid:3300)(cid:3301)(cid:4667)(cid:4666)(cid:3045)(cid:3299)(cid:3301)(cid:4667) If the first order partial is lower than the value of the zero-order coefficient, it is a direct relationship between variables x and y. Our next step in statistical analysis would probably be to discard this control variable and select another. The more the bivariate relationship retains its strength across a series of controls for third variables (zs), the stronger the evidence for a direct relationship x and y.
