Nonrandom sampling in human genetics: familial correlations

Area of Science:

• Human genetics
• Biostatistics
• Quantitative genetics

Background:

• Genetic studies often use nonrandom sampling, selecting families with affected individuals or extreme phenotypes.
• This biased sampling can distort genetic analyses, particularly for quantitative traits and diseases defined by cut-points.
• Implications of nonrandom sampling on genetic inference are often underestimated.

Purpose of the Study:

• To investigate the impact of nonrandom sampling on genetic analyses, specifically path analysis.
• To evaluate existing methods for correcting nonrandom sampling bias.
• To develop and validate a novel statistical approach to address these biases.

Main Methods:

• Path analysis was used to model genetic and environmental influences.
• Simulations were conducted to compare different sampling strategies and estimation methods.
• Method-of-moment estimators were developed for population parameters under nonrandom sampling.

Main Results:

• Nonrandom sampling leads to erroneous model specification and biased parameter estimates in path analysis.
• Conventional correction techniques (e.g., proband elimination, regression) also yield biased results.
• The proposed method-of-moment estimators demonstrated superior performance in simulations.

Conclusions:

• Inferences in genetic studies can be artifacts of sampling design rather than true biological effects.
• Existing correction methods are insufficient to mitigate biases from nonrandom sampling.
• Method-of-moment estimators provide a robust solution for accurate genetic parameter estimation with nonrandom samples.