Designs and Causality Flashcards
Simpsons Paradox
Simpsons Paradox is an extreme from of the fact that observed association can be misleading when there are lurking variables
–> conclusions that seem obvious when we look at aggregated data can become quite different when the data are examined in more detail
Causation
when you manipulate X —> Y changes
–> all other factors are held fixed and only X causes the change
Common response
When an observed association between X and Y is explained by a lurking variable Z
–> both X and Y change in resonse to Z
Causal relationship
- -> experimental research
- one variable directely influences another
1. unidirectional
2. bidirectional
Correlational research
–> correlation research
Changes in one variable accompany changes in the other variable, but there haven´t been enough tests to establish that either variable actually causes changes in the other
Conditions for inferring causal relationships
- Covariation
- Precedence - the hypothesized causal variable must reliably precede the effect variable
- Logical Mechanism - Plausability
Mediator variable
X does not directly influence Y ,but only indirectely via Z
–> Z is the mediator variable
Moderator variable
The relationship between X and Y differs according to the values of Z
–> Z acts as the moderator variable - likke a gate
Cross-sectional Design
Type of a correlational design
–> measurements are mad at the same point of time
Longitudinal Design
Type of a correlational design
–> measurements are madde at two or more different time points
Third variable/lurking variable problem
confounding
Correlation between variables could be due to a third variable that is not discovered yet
Directionality Problem
With only a correlational study it is impossible to determine unambiguously whether variable A causes variable B or vice versa
Independent Variable
The variable that is manipulated by the experimenter
Dependent Variable
The variable you observe and that gets influenced by the independent variable
Extraneous variable
Variable that may affect behaviour, but that is not of interest
Quasi Experiment
like an experiment but does not directly manipulate one variable. It tries to isolate a casual influence by selection rather than manipulation. Select cases in which X varies instead of varying x
Quasi independent variable
in experimental design, any of the personal attributes, traits, or behaviours that are inseparable from an individual and cannot reasonably be manipulated. These include gender, age, and ethnicity
Developmental
- Cross-sectional Design –> selecting data from a population at a specific point of time
- -> participants from each age group - Longitudinal Design –> a single group of participants is followed over a time period
Generation effect
Influences of different generations on experience which become confounded with the effects of age
–> in cross-sectional design
Cross-generational effect
-developmental design
conclusion drawn from examinated generation may not apply to another generation
Multiple observation effect
-developmental design
- Improvement of the test over time may be related to increasing experience and not age –> carryover effect
- Other factors tend to arise and become confounded with age –> history effect
Outcome research
investigates the effectiveness of treatment
–> goal: to determine whether a treatment produces a substantial or clinically significant effect
Process research
attempts to identify the active components of the treatment
