STAT201-Lecture 6-Confirmatory Factor Analysis
STAT201-Lecture 6-Confirmatory Factor Analysis
Confirmatory Factor Analysis (CFA) is a statistical technique used to test whether the data fit a
hypothesized measurement model. It is widely used in social sciences, psychology, and other
fields to validate the structure of latent constructs and their relationship with observed variables.
1. Purpose of CFA
• Validate Measurement Models: Test how well observed variables represent latent
constructs.
• Hypothesis Testing: Confirm whether a predefined factor structure fits the data.
• Evaluate Construct Validity: Ensure constructs are measured accurately.
2. Key Concepts
• Latent Variables: Unobservable constructs measured indirectly through observed
variables (indicators).
• Observed Variables: Measured variables (e.g., survey items or test scores).
• Factor Loadings: The strength of the relationship between latent variables and observed
variables.
• Measurement Model: A predefined structure specifying the relationships between latent
variables and their indicators.
Specify the hypothesized relationships between latent variables and observed variables. For
example:
• Latent Variable 1 (e.g., "Job Satisfaction") might be represented by indicators Q1, Q2,
and Q3.
• Latent Variable 2 (e.g., "Organizational Commitment") might be represented by
indicators Q4, Q5, and Q6.
Evaluate how well the model fits the data using fit indices:
• Chi-Square Test (χ²): Assesses overall fit. A non-significant χ² indicates good fit (but
sensitive to sample size).
• Root Mean Square Error of Approximation (RMSEA): Values < 0.05 indicate good
fit.
• Comparative Fit Index (CFI): Values > 0.90 (preferably > 0.95) indicate good fit.
• Standardized Root Mean Square Residual (SRMR): Values < 0.08 indicate good fit.
• Factor Loadings: Ensure they are statistically significant and > 0.5.
• Covariances: Check relationships between latent variables.
• Modification Indices: Identify areas for potential model improvement.
4. Example of CFA
Suppose a researcher hypothesizes a two-factor model for employee engagement:
Hypothetical Output
• Factor Loadings:
o Job Satisfaction: Q1 (0.75), Q2 (0.80), Q3 (0.70)
o Organizational Commitment: Q4 (0.85), Q5 (0.78), Q6 (0.72)
• Fit Indices:
o RMSEA = 0.04
o CFI = 0.96
o SRMR = 0.06
5. Assumptions of CFA
• Normality: Data should approximate a normal distribution.
• Linearity: Relationships between observed variables and latent constructs should be
linear.
• Adequate Sample Size: Minimum of 5-10 cases per parameter.
• Independence: Observations should be independent.
Issue 3: Multicollinearity
• Solution: Combine highly correlated latent variables or use exploratory factor analysis
(EFA).
Disadvantages