Imputation of missing ages in pedigree data.

Area of Science:

• Genetics and Bioinformatics
• Epidemiology
• Statistical Genetics

Background:

• Missing age at disease occurrence is common in human pedigree data.
• Imputation methods are used to address missing temporal data, but their performance in family studies is not well understood.
• The agreement between imputed and actual temporal data and its impact on inferences require further investigation.

Purpose of the Study:

• To evaluate the performance of five imputation methods for temporal data in human pedigree studies.
• To assess the agreement between imputed and actual disease diagnosis years.
• To determine the effect of imputation methods on statistical inferences, specifically regression coefficients.

Main Methods:

• Two evaluations were conducted using a prostate cancer family study dataset.
• Five imputation methods (mean substitution, regression, multiple imputation) were tested.
• Missing disease diagnosis years were randomly deleted and then imputed; agreement with actual values and regression coefficients was assessed.

Main Results:

• A specific regression method, utilizing birth year and last observation year of affected men, demonstrated optimal or near-optimal performance.
• The multiple imputation version of this regression method also yielded strong results.
• Agreement between imputed and actual diagnosis years was highest with this recommended regression approach.

Conclusions:

• The proposed regression imputation method is highly effective for handling missing diagnosis years in pedigree data.
• This method provides reliable temporal data for genetic and epidemiological research.
• Researchers can confidently use this imputation technique for improved inferences in family-based studies.