Structural Equation Modeling With Many Variables: A Systematic Review of Issues and Developments.

Area of Science:

• Social, behavioral, and health sciences research.
• Statistical modeling and data analysis.

Background:

• Survey data frequently involve numerous variables (p).
• Structural equation modeling (SEM) is a standard technique for analyzing relationships in such data.
• Traditional SEM requires substantial sample sizes (N) and can be problematic with small N or large p.

Purpose of the Study:

• To review existing issues and emerging solutions for SEM when dealing with small sample sizes (N), particularly when the number of variables (p) is large.
• To examine traditional sample size recommendations and contrast them with recent methodological developments.

Main Methods:

• Review of literature on parameter estimation methods for SEM with small N and large p.
• Analysis of test statistics for overall model evaluation in challenging SEM scenarios.
• Investigation of methods for reliable standard errors in SEM parameter significance testing.
• Examination of sample size (N) requirements and recent advancements.

Main Results:

• Conventional SEM methods often necessitate larger sample sizes (N) than anticipated.
• Newer methodologies and developments can significantly decrease the required sample size (N) for SEM analyses.
• Advances address challenges in parameter estimation, model evaluation, and significance testing for SEM with many variables (p).

Conclusions:

• Researchers using SEM with small sample sizes (N) and many variables (p) should be aware of current methodological advancements.
• New techniques offer solutions to previously identified problems, potentially making SEM more accessible in data-rich but participant-limited studies.
• Further methodological investigation is needed to address the complexities of SEM with big data (large p).