Internal/External Validity - Differential Selection

Vocabulary flashcards covering key terms and concepts related to differential selection and internal validity from Page 1 notes.

Differential selection (AKA Differential Assignment; Selection Bias)

A threat to internal validity that occurs when groups differ at the start of a study due to non-random assignment; these pre-existing differences may account for posttest outcomes.

Internal validity

The extent to which observed effects can be attributed to the treatment rather than other factors; differential selection threatens it.

Quasi-experimental design

A study design without random assignment, where groups are formed based on pre-existing conditions; higher risk of baseline differences.

Random assignment

Process that gives participants an equal chance of being placed in any group, helping ensure equivalent groups at baseline.

How does differential selection threaten internal validity?

If groups differ in key ways before the treatment (e.g., motivation, SES, symptom severity), it's unclear whether group differences at posttest are due to the treatment or those pre-existing characteristics.

How can researchers reduce differential selection?

• Use random assignment (if possible)

• Match participants across groups on key variables

• Use statistical controls (e.g., ANCOVA) - Measure and report baseline differences

Baseline differences

Pre-treatment disparities in characteristics between groups that can confound treatment effects.

Pre-existing groups

Groups defined by existing conditions (e.g., schools, clinics, communities) before assignment.

Differential selection vs. maturation

Differential selection refers to initial unequal groups; maturation refers to natural changes over time regardless of group.

EPPP-style clue for differential selection?

Vignette mentions: - No random assignment - Pre-existing groups (e.g., two schools, clinics, classrooms) - Groups differ in demographics or motivation

True experiment vs. quasi-experiment – Which is more at risk for differential selection?

Quasi-experiment, because it does not use random assignment, increasing the risk of group differences at baseline.

Quasi-experiment risk (differential selection)

In quasi-experiments, lack of random assignment increases the risk that group differences at baseline drive results.