Types of Third Variables Flashcards
Confounding Variable
A variable is associated with both the exposure and the outcome, but does NOT lie on the causative pathway
Confounding can account for what
Can account for all or part of an apparent association between an exposure and an outcome
What must be considered for confounding variables?
Prevent distortion of results
-Confounder is in the middle
Moderator Variable
A variable that alters the strength and/or direction of the relationship between the independent variable and the dependent such that the relationship between the independent variable and the dependent variable depends on the values of the moderator
Moderator can account for what
The exposure has a different effect on the outcome depending on the different subgroups (moderator) incorporated
What must be considered for moderator variables?
Stratification
-Act on the exposure
Mediator Variable
An intermediate or intervening variable in a causal chain relating an independent and a dependent variable
Mediator can account for what
Independent variable leads to changes on a mediator, which, in turn causes changes on a dependent variable
What must be considered for mediator variables?
Causative Pathway
-Mediator acts directly on the outcome
When is an RCT NOT feasible?
- Observational studies using administrative medical claims data
- Clinical studies with ethical prohibition on randomization
- Survey based research where respondents volunteer to participate but may not answer all survey items
What is a Natural Experiment?
- Partially uncontrollable or controllable studies
- Flaws do NOT average out
What is Regression Analysis?
A FAMILY of stat techniques that assess differences across test and control groups while accounting for factors that do NOT average out (bias)
Can confounders, moderators, and mediators exist simultaneously?
YES, they can exist independently or simultaneously
