Gain in efficiency from using generalized least squares in the Haseman-Elston test

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Linkage analysis is crucial for identifying genes associated with traits.
• Haseman-Elston (H-E) regression methods are commonly used for linkage testing.
• Existing modified H-E methods (OLS, WLS, GEE) often incorrectly assume independence between sib pairs within families.

Purpose of the Study:

• To develop and evaluate a generalized least squares (GLS) estimator for Haseman-Elston linkage analysis.
• To correctly account for correlations between sib pairs within families.
• To compare the efficiency of the proposed GLS estimator with existing WLS/OLS methods.

Main Methods:

• Implemented a generalized least squares (GLS) estimator incorporating sib pair correlations.
• Determined the null variance of the GLS estimator and compared it to WLS/OLS estimators.
• Analyzed the efficiency gains of GLS over WLS/OLS under the null hypothesis of no linkage.

Main Results:

• The null variance of the WLS/OLS estimate can be larger or smaller than previously calculated.
• The GLS estimator provides a more accurate assessment of the regression coefficient by accounting for sib pair correlations.
• Efficiency gains of approximately 11% (3 siblings/family) and 25% (4 siblings/family) were observed with GLS in large multifamily studies.

Conclusions:

• Correctly specifying correlations between sib pairs enhances the accuracy of linkage analysis.
• The proposed GLS method offers improved statistical power for detecting genetic linkage compared to standard WLS/OLS approaches.
• This work provides a more robust statistical framework for genetic linkage studies.