Single missing data in the N = 375 sample was imputed using the EM-algorithm [ 17– 20].
Fifteen (10 training, 5 validation) samples were imputed using multivariate imputation chain equations.
Samples were imputed using the 1000 genomes CEU data as the phased reference.
Missing data for efficacy variables were imputed using the last-observation-carried-forward (LOCF) method.
Scores for missing items were imputed using the average scale score from the completed items.
The estimates for the remaining countries were imputed using analogy with countries in the same region.
For all survival analyses, missing data were imputed using multiple imputations using chained equations [ 17, 18].
Missing values in sampling time of day for Eden cohort (n = 25) were imputed using linear regression adjusted for date and sampling season, parity, education level, occupation, active smoking and center.
Illumina plattforms were used for genotyping and missing SNPs were imputed using the 1000 Genomes data.
To maintain consistent sample sizes across markers, these missing values were imputed using the median of the sample.
