Lecture 2: Causality, Populations, Time

Directed Acyclic Graphs (DAGs)

Diagrams made of nodes (variables) and directed edges (arrows) showing hypothesized causal relationships.

Confounder

A variable that influences both the exposure and the outcome but is not in the causal pathway between them.

Backdoor Criterion

A variable (or set of variables) is a sufficient adjustment set if, when conditioned, all backdoor paths between exposure and outcome are blocked.

Collider

A variable that is influenced by two other variables; controlling for it can create spurious associations.

Mediator

A variable that lies on the causal pathway between an exposure and outcome, affecting the direct relationship.

Hill's Criteria

A set of nine considerations to guide whether an observed association is likely causal.

Strength of Association

Stronger associations are less likely to be confounded, indicating a more direct causal connection.

Statistical Significance

A measure that indicates whether a result is likely due to chance; does not confirm causation.

Causal Pathway

The sequence of events leading from an exposure to a disease outcome.

Causal Inference

The process of drawing conclusions about causal relationships based on observed data.

Strength of Association (Hill's Criteria)

A stronger association between a potential cause and an effect makes a causal relationship more likely; weaker associations are more susceptible to confounding.

Consistency (Hill's Criteria)

If the association is observed repeatedly by different people, in different places, circumstances, and times, it strengthens the likelihood of a causal relationship.

Specificity (Hill's Criteria)

While not always true, if a single exposure leads to a single disease outcome, it increases the probability of a causal relationship.

Temporality (Hill's Criteria)

The cause must precede the effect in time. This is the only absolutely essential criterion for causality.

Biological Gradient / Dose-Response (Hill's Criteria)

Increased exposure to the potential cause should correspond to an increased effect (e.g., higher dose, higher risk).

Plausibility (Hill's Criteria)

There should be a plausible biological mechanism or theoretical basis explaining how the cause leads to the effect.

Coherence (Hill's Criteria)

The observed association should be consistent with existing knowledge and theories regarding the natural history and biology of the disease.

Experiment (Hill's Criteria)

Evidence from experimental studies (e.g., randomized controlled trials) can provide strong support for a causal relationship.

Analogy (Hill's Criteria)

If similar causal relationships have already been established for similar exposures or outcomes, it strengthens the case for a new causal relationship.

Causation

A relationship where an exposure leads to a change in a disease or outcome.

Target Population

The specific group of individuals about whom inferences are to be made in a research study.

Source Population

The group of individuals expected to have the same exposure-disease relationship as the target population.

Study Population

The group of individuals under observation in the study, expected to have the same exposure-disease association as the source population.

Confounding Factor

An extraneous factor that distorts the apparent effect of the exposure on the disease.

Representativeness

How accurately the sample represents the entire population.

Descriptive Research

Research aimed at describing the occurrence of factors or disease in a population at a defined time.

Directed Acyclic Graphs (DAGs)

A graphical representation of causal pathways that helps identify causal relationships in epidemiological research.

Causal Pathway

A sequence of events or processes that lead from an exposure to an outcome.