Session 2 Flashcards
Association
The co-occurrence of two factors in a population. When two variables are associated, knowing the value of one variable helps you predict the value of the other.
- DIRECTIONALITY does not matter in a DAG when looking at association
Berkson’s Bias
A type of collider bias that arises in the context of selecting individuals for a study from treatment settings.
Collider
A variable on a specified path that is entered and exited through an arrowhead; a common effect of two variables in a DAG.
Collider Bias
The bias that arises from conditioning on (controlling for) a collider.
D-Separation
When two variables in a DAG are not connected by an open backdoor path.
D-Separation Rules
A set of rules to d-separate an exposure and disease in a DAG from open backdoor paths.
DAG
Directed Acyclic Graph. A graphical method for depicting hypothesized causal relationships between variables and deducing the statistical associations implied by these causal relationships.
By directed, we mean that the arrows point in only one direction because they indicate causation.
By acyclic, we mean that no variable can cause itself directly or through any number of other variables.
Direct Effect
In the context of DAGs, an arrow connecting two variables that is not intercepted by another variable on the path; a causal relationship with no articulated mediator.
Directed Path
A series of arrows all pointed in the same direction (a directed pathway) indicates indirect causes.
Edges
The paths created by the arrows connecting variables are called edges.
Family Relationships in a DAG
Parent: A variable that is a direct cause of another variable in a DAG.
Child: A variable that is the direct consequence of another variable in a DAG.
Ancestor: When distinguished from a parent - it is an indirect cause of another variable in a DAG; it is a direct or an indirect cause of another variable; parents are ancestors
Descendant: When distinguished from a child - it is the effect of an indirect causes; it is the effect of a direct or an indirect cause; children are descendants.
Indirect Effect
In the context of DAGs, an arrow connecting two variables that is intercepted by another variable on the path; a causal relationship with an articulated mediator.
Node
The variables in a DAG.
Path
A sequence of lines connecting two variables.
Blocked Path: A path that is intercepted by a controlled variable or an uncontrolled collider.
Unblocked Path: A path that is not intercepted by a controlled variable or an uncontrolled collider.
Backdoor Path: A path that connects two variables, where the path does not emanate from one of the variables connected.
Frontdoor Path: A path that connects two variables, where the path does not emanate from one of the variables being connected.
Prospective Cohort Study
A study where participants are free of the outcome at the beginning of the study, the exposure is recorded, and the participants are followed over time to see who develops the disease.
