Design considerations in the estimation of intraclass correlation

Area of Science:

• Biostatistics
• Quantitative Genetics
• Human Genetics

Background:

• Estimating the intraclass correlation coefficient (ICC) is crucial for understanding familial aggregation of traits.
• Family study designs are commonly used in genetic epidemiology and behavioral genetics.
• The efficiency of ICC estimation depends on the sampling strategy within families.

Purpose of the Study:

• To compare the accuracy of balanced versus unbalanced family study designs for estimating the ICC.
• To evaluate the impact of family size variation on ICC estimation precision.
• To provide guidance on optimal family study design for genetic research.

Main Methods:

• Simulation studies were conducted to compare two estimators of the ICC.
• Balanced designs (equal number of individuals per family) were contrasted with unbalanced 'natural' designs.
• The performance of designs was assessed based on the accuracy of ICC estimation.

Main Results:

• Balanced family study designs generally offer slightly higher accuracy for ICC estimation.
• The advantage of balanced designs diminishes significantly when the number of families exceeds 50.
• The choice of ICC estimator can influence the relative performance of designs.

Conclusions:

• Balanced designs are often preferable for family studies aiming to estimate ICC, particularly with fewer than 50 families.
• For large sample sizes (k > 50), the benefits of a strictly balanced design are marginal.
• Researchers should consider sample size and potential gains in precision when choosing between balanced and unbalanced designs.