Research Methods - Design Pt3 - Between-Subjects and Within-Subjects Designs

Experimental Design: Between Subjects vs. Within Subjects

Overview of Experimental Designs

• The video discusses two primary experimental designs: between subjects and within subjects. Each design has distinct methodologies for comparing groups or conditions.

Between Subjects Design

• In a between subjects design, different groups of participants are assigned to different conditions, allowing for independent comparisons. For example, one group may be assigned a vegetarian diet while another is assigned a meat diet.

• This design controls for pre-existing differences by randomly assigning participants to conditions, which enhances the validity of the findings compared to merely observing existing groups (e.g., vegetarians vs. meat-eaters).

• An example research question could be whether people spend differently before and after taking a budgeting course; this would typically involve measuring separate groups rather than the same individuals over time.

Implications of Between Subjects Design

• The results from between subjects experiments can be analyzed using an independent samples T-test, which compares means from two distinct groups to infer population characteristics. This statistical tool is essential for determining significant differences in outcomes across independent samples.

• Even when not conducting true experiments (e.g., comparing cat lovers vs. dog lovers), an independent samples T-test remains applicable as it evaluates differences between any two distinct groups regardless of data type (experimental or correlational).

Within Subjects Design

• A within subjects design involves testing the same group of participants under different conditions, such as measuring their performance before and after treatment or comparing scores on sleep-deprived versus well-rested days. This approach allows for paired comparisons since each participant serves as their own control.

Understanding Within-Subjects Design and Dependent Samples T-Test

Overview of Within-Subjects Design

• Within-subjects design involves comparing two sets of related or paired scores, such as before versus after treatment.

• The dependent samples T-test (also known as the related samples T-test) is used to determine statistically significant differences between these paired scores.

Mechanics of the Dependent Samples T-Test

• This test evaluates the mean difference by calculating the average change from before to after treatment for each participant.

• Each participant's scores are aligned in rows, allowing for direct comparison and calculation of individual difference scores.

• The average of these difference scores is then analyzed to assess statistical significance.

Applications and Examples

• The dependent samples T-test can also be applied to matched or paired data, such as twin studies or spouse comparisons.

• An example includes assessing husbands' perceptions of their chore contributions compared to their wives' perceptions.

Advantages of Within-Subjects Design

• A key benefit is that each participant serves as their own control group, reducing variability due to individual differences.

• This design allows for more precise measurement since all participants experience every condition being tested.

Limitations: Carryover Effects

• Unlike between-subject designs where different individuals are assigned to conditions, within-subject designs may suffer from carryover effects.

Understanding Carryover Effects in Experimental Design

Types of Carryover Effects

• A practice effect is a type of carryover effect where participants improve on memory tasks due to prior exposure to similar tasks. This raises questions about the order of conditions in experiments.

• The fatigue effect occurs when earlier conditions tire or demotivate participants, potentially leading to poorer performance in subsequent conditions. This highlights the need for careful consideration of condition ordering.

• A contrast effect can arise when experiences from an earlier condition influence responses in a later one, such as biasing reactions based on previous stimuli (e.g., beautiful vs. ugly faces).

Addressing Carryover Effects

• All carryover effects pose challenges in within-subject experiments where participants undergo multiple conditions sequentially, making it crucial to find ways to minimize their impact.

• The primary solution for mitigating carryover effects is counterbalancing, which involves varying the order of conditions among different participants to balance out potential biases and effects.

Implementing Counterbalancing

• In studies with more than two conditions, it's essential to ensure that all possible orderings are represented evenly across participants, thus reducing the risk of skewed results due to specific sequences.

• Researchers can analyze data post-experiment to quantify any carryover effects by comparing scores based on the sequence of task completion (e.g., hard after medium vs. hard after easy).

Counterbalancing Techniques

• The ideal method for counterbalancing is complete counterbalancing, which assigns participants equally across all possible orders; however, this requires a large number of participants as complexity increases.

• An alternative approach called a Latin Square design offers efficiency while still providing significant benefits by balancing out orderings without needing as many subjects.

Comparing Experimental Designs

• Between-subject designs are straightforward and avoid carryover effects since each participant only engages with one condition but require more subjects for adequate statistical power.

Research Design Considerations in Psychology

Understanding Carryover Effects

• The primary concern with research designs is the carryover effects, which can influence results across different conditions.

• Counterbalancing is a method used to minimize carryover effects, although it may not completely eliminate them.

Multiple Approaches to Research Questions

• A single research question can often be addressed through various methodologies, including both between-subjects and within-subjects designs.

• Confidence in the validity of an effect increases when studied through multiple designs and operational definitions.