Testing the equality of twin correlations with multinomial outcomes

Area of Science:

• Biostatistics
• Behavioral Genetics
• Quantitative Genetics

Background:

• Twin studies are crucial for dissecting genetic and environmental contributions to human traits.
• Correlation analysis is a common method to assess twin similarity in these studies.
• Evaluating correlations for multinomial outcomes requires robust statistical approaches.

Purpose of the Study:

• To compare the performance of three statistical methods for testing the equality of twin correlations with multinomial outcome variables.
• To identify the most reliable method under varying sample sizes and trait prevalences.
• To provide practical guidance for analyzing twin data in genetic and environmental research.

Main Methods:

• Likelihood ratio test utilizing a Dirichlet-multinomial distribution.
• Method based on the estimated large sample variance of the correlation.
• Chi-squared goodness-of-fit test.

Main Results:

• All three methods demonstrate similar performance with large twin pair samples (>100) and non-extreme trait prevalences.
• The chi-squared goodness-of-fit test is preferable when sample sizes are small or trait prevalence is extreme.
• The study illustrates these methods using data from a smoking behavior twin study.

Conclusions:

• The choice of method for testing twin correlations with multinomial outcomes depends on sample size and trait prevalence.
• The chi-squared goodness-of-fit test offers a reliable alternative, particularly in challenging data scenarios.
• Accurate statistical analysis in twin studies is essential for understanding human trait etiology.