Statistical validity for testing associations between genetic markers and quantitative traits in family data

Area of Science:

• Genetics
• Statistical Genetics
• Biostatistics

Background:

• Genetic analysis often examines associations between traits and genetic markers using pedigree data.
• Standard statistical methods assuming independence are unsuitable due to inherent statistical dependencies among pedigree members.
• Quantitative traits require specialized models that account for these complex familial relationships.

Purpose of the Study:

• To assess the statistical validity of the George and Elston regression model for finite samples.
• To determine the minimum pedigree size required for reliable inferences using the asymptotic likelihood ratio test.
• To evaluate the relationship between the variance of the dependent variable and the discrepancy in chi-square distributions.

Main Methods:

• Utilized a regression model incorporating a correlation structure for statistical dependence among pedigree members.
• Employed an asymptotic likelihood ratio test for association analysis.
• Assessed statistical validity by measuring the discrepancy between empirical and theoretical chi-square distributions for finite samples.

Main Results:

• The variance of the mean of the dependent variable is linked to the discrepancy between empirical and theoretical chi-square distributions.
• This relationship can inform the adequacy of pedigree size for valid statistical inferences.
• A multi-generational pedigree of approximately 200 individuals is often sufficient for valid results.

Conclusions:

• The George and Elston regression model provides a valid approach for analyzing associations between continuous traits and genetic markers in pedigrees.
• Pedigree size is a critical factor in the reliability of asymptotic tests.
• Approximately 200 individuals in a multi-generational pedigree generally suffice for valid statistical inferences.