Within-Subjects Designs Lecture Notes

Vocabulary flashcards covering the definitions, benefits, issues, solutions, and analysis methods for within-subjects research designs.

Within-subjects designs

Designs that involve measuring participants on a dependent variable multiple times.

Pretest-posttest designs

Designs involving measuring some outcome before and after an intervention or manipulation, typically the simplest kind of within-subjects design.

Repeated measures designs

Designs that involve exposing participants to each level of the independent variable and measuring outcomes after each exposure.

Longitudinal within-subjects designs

Designs that compare participants to themselves at multiple time points to address questions of change over time on a dependent variable.

Efficient Hypothesis Testing

An advantage of within-subjects designs where each participant is in every condition, allowing multiple hypotheses to be tested with a single sample (e.g., an IV with $3$ levels needs only $1$ sample).

Participants as Own Controls

A feature of within-subjects designs that significantly reduces error by directly accounting for individual differences instead of relying on random assignment.

Independence of Data

A major statistical assumption that each data point comes from different people, which is violated by within-subjects designs by definition.

Learning effects

An issue where exposure to a manipulation or research materials (e.g., surveys) affects future responses or loses reliability, such as in the testing effect.

Maturation effects

A threat to inference where participants change over time for reasons other than the experimental manipulation.

Attrition effects

A threat to inference occurring when participants leave a study, resulting in the loss of their data from every experimental condition.

Order effects

Problems caused by the specific sequence in which participants are exposed to research materials.

Practice effects

A specific order effect where participants change their behavior or responses due to familiarity with the measures used.

Fatigue effects

A specific order effect where participants become tired or bored over the course of a study, introducing error into measurement.

Carryover effects

A specific order effect where earlier manipulations affect the responses to or engagement with later manipulations.

Sensitization effects

A specific order effect where exposure to study materials leads participants to try and guess the research hypotheses, which can affect their behavior.

Oversampling

The practice of collecting more participants than necessary for a target level of power, often by a specified percentage like $50-75\%$ or $2-4\times$ to compensate for attrition.

Counterbalancing

The procedure of presenting measures and manipulations in all possible orders to different participants to wash out order effects.

Factorial Ordering

A limitation of counterbalancing where the sample size explodes because the number of possible orders scales factorially with the number of options (e.g., $4$ manipulations yields $4! = 24$ orderings).

Latin square design

A design that controls for position effects by ensuring each measure or manipulation appears in each possible position only once, requiring fewer orders than full counterbalancing.

Paired-samples t-test

Also called a dependent samples t-test, it is used to compare two sets of dependent scores by computing the difference set for each participant.

Repeated measures ANOVA

Also called a within-subjects ANOVA, it tests whether there is at least one difference between multiple sets of scores (time points or manipulations).