1/18
This set covers key terms and concepts from Chapter 10 regarding multivariate relationships, causal criteria, and types of statistical associations.
Multivariate Analysis
Statistical methods used to include other variables in an analysis that might influence the association between two primary variables.
Causality
A central scientific concept usually characterized by asymmetry, where one variable influences another but not vice versa.
Association
The first criterion for causality; it requires showing that as X changes, the distribution of Y changes in some way.
Time Order
The second criterion for causality requiring that the cause precedes the effect in time.
Elimination of Alternative Explanations
The third and most difficult criterion for causality, requiring researchers to show that no other variable can explain the observed association.
Statistical Control
A tool for studying how an association changes after removing the influence of other variables by grouping observations with equal or similar values on those variables.
Control Variable
A variable whose value is held constant to remove its influence on the association between two other variables.
Partial Tables
Separate contingency tables that display the relationships between two categorical variables within fixed levels of a control variable.
Lurking Variable
A variable not measured in a study (or perhaps unknown to the researchers) that influences the association of interest.
Spurious Association
An association between X and Y that results from their common relationship with a third variable X2, and disappears when X2 is controlled.
Chain Relationship
A multivariate structure where X1 affects X2, which in turn affects Y, making X1 an indirect cause of Y.
Intervening Variable
A variable (also called a mediator variable) that occurs between an explanatory variable and a response variable in a causal chain.
Multiple Causes
A situation where a response variable (Y) is influenced by more than one explanatory variable (X1, X2, etc.).
Independent Causes
Explanatory variables that are statistically independent of each other while each having a causal effect on the response variable.
Suppressor Variable
A third variable that, when not controlled, hides an association between two other variables; the association only appears once this variable is controlled.
Statistical Interaction
Exists when the true effect of one predictor on Y changes as the value of another predictor changes.
Simpson's Paradox
A situation where the associations in partial tables have the opposite direction as the association in the bivariate table.
Confounding
Occurs when two explanatory variables are associated with each other and both affect the response, making it difficult to determine the true causal source.
Measure of Partial Association
A summary measure that describes the strength of an association between two variables conditional on a control variable.