Chapter 7
Research Methods: A Process of Inquiry - Chapter 7 Correlational and Differential Research
Announcements
Exam: Next Wednesday, 3/4
Lab #7: Due 2/27
Lab #8: Due 3/6
Learning Objectives
Compare and contrast correlational and differential research methods
Determine relationships between variables using correlation
Describe how to conduct differential research
Describe the limitations of interpreting correlational and differential research
Explain how low-constraint research helps provide information when ethical constraints limit experimentation
Big Topics
Correlational research methods
Differential research methods
Limitations of both methods
Ethical principles relating to research methodology
Correlational Research
Purpose: Quantifies the strength of the relationship among two or more variables.
Value:
Correlations can be used for prediction.
Provide evidence that is consistent or inconsistent with a theory.
Correlations cannot establish causation (e.g., the common belief that "very bright children are socially challenged").
Understanding Correlation and Causation
If A and B are correlated, several scenarios exist:
A could cause B.
B could cause A.
Another variable could cause both A and B.
Care must be taken in drawing conclusions from correlations.
Differential Research
Definition: Compares two or more preexisting groups.
Similarities: Related to both correlational and experimental research in format.
Comparison of Groups: Like experimental research, but here variables are measured and not manipulated.
Cross-Sectional Design: Used in developmental research, analyzing different groups at a single point in time.
Cross-Sectional and Longitudinal Research
Cross-Sectional Designs: Offer quicker results by testing various age groups simultaneously but may encounter cohort effects.
Cohort Effects: Defined as shared life experiences of people of a given age leading to similar behavior compared to other ages.
Longitudinal Designs: Time-series studies with multiple measurements before and after an intervention.
Time Series Designs
Purpose: Hybrid of cross-sectional and longitudinal designs.
Method: Involves successive independent samples over time by asking the same questions to new samples periodically for longitudinal trends.
Examples: Used for measuring consumer confidence, ratings of Congress, and public opinion polls.
Artifacts and Confounding Variables
Definition of Confounding: Occurs when two variables vary together. Effective study design requires them to vary independently, which often entails controlling for variables.
If not controlled, the results may lead to artifactual findings.
Standardization of Procedures: Important to ensure comparable measurements across groups.
Example in Practice: Studying schizophrenia with comparison samples matched on age, education level, and psychiatric hospitalization.
Example of Confounding Variables
Hypothesis: Drinking coffee improves memory recall.
Two groups formed for testing:
No Coffee Group: Tested at 5 AM.
Coffee Group: Tested at 2 PM.
Correlational Versus Differential Research
Both types of research involve measurement without manipulation, thus neither can establish causation.
Differential Research: Considered higher constraint as the researcher can choose comparison groups to control confounding variables, which helps provide stronger evidence for hypotheses.
Choosing Research Methods
Correlational Method: Appropriate when the primary interest is knowing the strength of relationships for predictive purposes; often used for interpreting findings.
Differential Research: Utilized when manipulation of an independent variable is impractical, impossible, or unethical, leading to comparisons among preexisting groups.
Steps in Conducting Correlational Research
Develop the problem statement.
Measure the variables.
Obtain a sample.
Analyze the data.
Interpret the results.
Often embedded as secondary analyses in broader studies.
Developing a Problem Statement
Formulate questions such as: “What is the relationship between variable X and variable Y?”
Useful for correlating available demographic variables or multiple measures in research.
Measuring the Variables
Use measures that are reliable and valid while controlling for potential biases:
Experimenter Expectancy: Researchers may see what they expect to see.
Experimenter Reactivity: Researchers might unconsciously influence participants.
Measurement Reactivity: Participants may respond differently due to awareness of observation.
Controlling for Bias in Measurement
To counter these effects, researchers are recommended to:
Use objective measures.
Minimize direct contact to reduce experimenter reactivity.
Employ filler items to distract participants or use unobtrusive measures.
Space out separate measurements over time to prevent interaction.
Sampling Considerations
A representative sample enhances generalizability:
Examine if the observed relationship holds across subpopulations.
Use moderator variables, which adjust the relationship between different variables; examples include gender, ethnicity, and culture.
Analyzing Correlational Data
Correlation coefficients can range from −1.00 to +1.00.
Size: Indicates strength of the relationship.
Sign: Indicates direction of the relationship.
Types of correlations:
Pearson Product-Moment Correlation: Assesses linear relationships.
Spearman Rank-Order Correlation: Non-parametric measure for ordinal data.
Phi Coefficient: For binary variables.
Advanced techniques include multiple correlation, canonical correlation, partial correlation, and path analysis.
Interpreting Correlational Data
Analyze the size and sign of the correlation to infer strength and direction.
Determine significance of correlation against zero:
Assess if $p < \alpha$ (alpha).
Coefficient of Determination: The square of the correlation coefficient, indicating the proportion of variance accounted for. Example:
If $r = 0.50$, then $r^2 = 0.25$ indicates that 25% of variance is accounted for.
Doing Differential Research: Steps
Develop the problem statement.
Measure the variables.
Select appropriate control groups.
Obtain a sample.
Analyze the data.
Interpret the results.
Problem Statement in Differential Research
Sample statement: “Does Group A differ from Group B?”
Focus on selecting groups that are theoretically interesting for comparison.
Ensure groups differ only on the specific variable of interest if possible.
Measuring Variables in Differential Research
Dependent variable: Usually continuous, but can also be categorical.
Independent variable: Generally categorical but can also be a continuous variable converted into categories.
Emphasis on clear operational definitions.
Selecting Control Groups
Control groups are essential to avoid confounding:
Confounding occurs if a variable affects dependent outcomes and groups differ on this variable.
Ideal to keep control groups as identical to experimental groups as possible, although this is often challenging.
Example Study Referenced
Blanchard et al. (2001): Analysis on “edonia” scores showcasing baseline and follow-up results to depict group behavioral differences.
Sampling of Participants
Representative sampling is vital for generalizability:
Situational factors (such as time of day) or participant dropout can influence study outcomes.
Analyzing Data in Differential Research
Employ similar analysis procedures as in experimental research, tailored to the number of groups and measurement levels:
For score data: use t-tests or ANOVA.
For ordinal data: apply Mann-Whitney U-test.
For nominal data: utilize Chi-square tests.
Interpreting Results in Differential Research
Null hypothesis (no group difference) is rejected if $p < \alpha$.
Caution is advised when drawing conclusions due to potential confounding variables.
Limitations of Correlational and Differential Research Methods
Challenges in determining causation: correlation does not imply causality.
Correlation scenarios:
A could cause B.
B could cause A.
C could influence both A and B.
Confounding variables may persist without experimental control.
Ethical Considerations in Research
Some causal hypotheses cannot ethically be tested in experimental designs:
Example: it is unethical to abuse children to determine effects on adult depression.
Ethical or practical constraints lead researchers to use differential and correlational research methods as alternatives.
Summary of Key Points
Both correlational and differential research methods focus on measuring relationships between variables.
Exercising caution is necessary when drawing causal inferences.
Selecting appropriate control groups in differential research can mitigate some, but not all, confounding variables.
Exam 2 Review Topics
List descriptive statistics defined in Chapter 5 and provide examples of their applications.
Discuss strengths and limitations of designs discussed in Chapters 6 and 7, organizing them into two clear lists.
Give examples of scenarios for employing correlational research and differential research.
Supplemental Slides and Resources
Website Resources:
Examples of Correlational Research
Examples of Differential Research
Computational Procedures for Correlations
Selecting Appropriate Statistical Analysis Procedures
Guide for Computing Various Statistical Tests
Student Study Guide/Lab Manual
Related Internet Sites
Determining Statistical Significance Flowchart
Select appropriate statistical test.
Compute the statistical test and its p-value.
If $p < \alpha$, conclude that population means are not equal; otherwise, they are equal.