Bivariate Correlations & Interrogating Association Claims
Research Methods
Chapter 8: Bivariate Correlations & Interrogating Association Claims
- Objective: Understand the role of measured variables in correlational studies.
- Key Distinction: Measured variables, rather than statistics, define a study as correlational.
- Validity of Association Claims:
- Critical to interrogate the validity of association claims.
- A correlational study supports an association claim but not a causal claim.
Research Scenario: Dr. Singh's Study
Description of the Study
- Dr. Singh's aim: Explore the relationship between self-esteem and students' willingness to participate in class.
- Methodology:
- Students fill out a self-report measure of self-esteem consisting of the following items:
- I feel confident.
- I am capable.
- I rarely feel good about myself.
- Response scale: 1 (strongly disagree) to 5 (strongly agree).
- Data collection: Two research assistants tally the number of times students raise their hands in class.
Research Questions
- Variables Identified:
- Self-esteem (measured).
- Willingness to talk in class (measured).
- Type of Variables:
- Both variables are measured, none are manipulated.
- Nature of Claim:
- Dr. Singh will make an association claim, not a causal claim.
- Validity Concerns:
- Important validities: Construct validity of self-esteem measure.
- Reliability Concerns:
- Significant reliability: Internal consistency of the self-esteem measure.
- Construct Validity Assessment:
- Method to assess: Compare self-esteem measure with established measures (e.g., Rosenberg Self-Esteem Scale).
- Desired evidence: High correlation with established measures, showing convergent validity.
- Question Type for Self-Esteem:
- The self-esteem measure uses declarative statements for responses.
- Concerns About Acquiescence:
- Dr. Singh could include opposite wording statements to balance responses.
- Observer Bias Reduction:
- Ensuring that research assistants are trained to be objective in observation,
- Participant Reactivity Reduction:
- Implementing indirect methods for counting participation, such as unaware observation.
Bivariate Correlation
Definition
- Bivariate Correlation: An association between exactly two measured variables.
- Association Claim: Describes the relationship between these two measured variables, asserting that two things are related without demonstrating causation.
Examples of Bivariate Correlation Data
Well-Being and Deep Talk Study (Mehl et al., 2010)
- ```plaintext
PERSON | SCORE ON WELL-BEING SCALE | PERCENTAGE OF CONVERSATIONS RATED AS DEEP TALK
A | 1.2 | 80
B | 0.3 | 52
C | -0.4 | 35
D | 1.4 | 42
ZZ | -1.6 | 16
- Note: Data are fabricated for illustration purposes.
- Correlation coefficient: r = .26.
## Sitting Time and MTL Thickness Study (Siddarth et al., 2018)
- ```plaintext
ADULT | TIME SPENT SITTING (HOURS) | MTL TOTAL THICKNESS (MM)
------------------------------------------------------------
A | 10 | 2.67
B | 11 | 2.56
C | 5 | 2.70
D | 7 | 2.51
JJ | 3 | 2.34
- Correlation coefficient: r = -.37.
Interrogating Association Claims
Key Validities to Assess
- Three types of validities: Construct validity, statistical validity, and external validity.
- Importance of internal validity: Correlational studies do not establish internal validity.
Construct Validity
- How well operationalized were the variables?
- Assessing validity:
- Measurement methods must capture the constructs of interest accurately.
Statistical Validity
Considerations:
- Strength and precision of the association:
- Effects of outliers on correlation coefficients.
- Replication of findings across multiple studies.
- Examination for restriction of range.
- Assessing if a zero association is curvilinear.
Three Criteria for Causation
1. Covariance
- Are the results showing that variable A changes with variable B?
- Example: Watching deeper conversations correlates with higher well-being.
2. Temporal Precedence
- Establishes that one variable must occur before the other.
- Important to resolve directionality issues.
3. Internal Validity
- Ensure that there are no plausible alternative explanations (third-variable problem) for the relationship between A and B.
External Validity
Generalization of Findings
- Can findings be generalized across various populations and contexts?
- Example: MTL Thickness and sitting study concluded using convenience sampling (may limit generalizability).
External Validity Concerns
- Considerations about the sample: Gender, age, how participants were recruited, etc.
Moderating Variables
Definition
- A variable that influences the strength or direction of the relationship between two other variables.
- Example: Relationship between attendance and team success moderated by local community ties.
Applications of Moderating Variables
- Assess whether certain groups have stronger relationships between A and B.
- Understanding how factors like parental discussion influence aggression related to media.