A limited dependent variable model for heritability estimation with non-random ascertained samples

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Asthma research often faces challenges with non-random sample selection.
• Volunteer bias can skew results when only a fraction of an initially selected sample participates in further studies.

Purpose of the Study:

• To present a statistical model for correcting sample selection bias.
• To enable representative population estimates from selected samples in family health studies.

Main Methods:

• Developed a linear regression model (DeFries-Fulker) incorporating sample selection correction.
• Utilized data from a Dutch family study on asthma and related symptoms.
• Compared the proposed model with raw likelihood estimation, multiple imputation, and sample weighting.

Main Results:

• The corrected model provides estimates representative of the original population.
• Demonstrated the model's efficacy using a real-world asthma family dataset.
• Results were compared against alternative statistical methods for bias correction.

Conclusions:

• The proposed limited dependent variable model effectively addresses sample selection bias.
• This approach enhances the generalizability of findings from selected study samples.
• The method is particularly useful in epidemiological studies with volunteer bias.