A Monte Carlo evaluation of three statistical methods used in path analysis.

Area of Science:

• Statistics
• Quantitative Psychology
• Behavioral Genetics

Background:

• Path analysis is crucial for family data, often employing maximum likelihood methods.
• Assumptions of multivariate normality and large sample sizes are typical for these methods.
• Existing methods differ in likelihood function specification, impacting results.

Purpose of the Study:

• To compare three maximum likelihood-based statistical methods for path analysis in family data.
• To evaluate the validity, efficiency, and bias of parameter estimates under various conditions.
• To assess the robustness of these methods against violations of normality and sample size assumptions.

Main Methods:

• A Monte Carlo simulation study with 1,000 replications per condition.
• Comparison of three methods differing in likelihood function specification.
• Investigation of effects of non-normality and small sample sizes on statistical properties.

Main Results:

• Method 3 demonstrated valid statistical inferences under multivariate normality and robustness to minor non-normality.
• Method 2 was robust to minor non-normality but sensitive to small sample sizes.
• Method 1 produced highly conservative test statistics across all simulated conditions.

Conclusions:

• Method 3 is recommended for family path analysis when multivariate normality holds.
• Researchers should consider sample size and normality when choosing between Method 2 and Method 3.
• Method 1 is generally too conservative for practical application in family studies.