We assumed data were missing at random, and we imputed missing age and depression scores at follow-up using a multivariate imputation algorithm ("mi impute mvn," in Stata version 11) using Markov Chain Monte Carlo.
We first imputed the missing data using mice (multiple imputation algorithm), based on 30 multiple imputed datasets.
Missing exposure data were imputed with a multiple imputation algorithm (SAS MI procedure, version 9.2; SAS Institute Inc., Cary, NC) based on the exposure data for all eligible persons (n = 2,222), creating 18 replicate imputed data sets.
Missing values were imputed using the k-NN imputation algorithm described by Troyanskaya and colleagues [ 20].
Missing values were imputed using the k-NN imputation algorithm (k = 10) described by Troyanskaya and coworkers [ 28].
Within the full conditional specification imputation algorithm, continuous variables were imputed using multiple linear regression, smoking status using multinomial regression and IMD quintile using ordered logistic regression.
Missing data were imputed by the k-nearest neighbors imputation algorithm [ 27].
A simple imputation algorithm was developed for the purpose of imputing the markers that were missing because of technical limitations with GBS data (not biologically missing).
Our model, therefore, implicitly assumes that an imputation algorithm will correctly jump between reference haplotypes when imputing a target.
Multiple imputation was performed by the MI and the MIANALYZE procedures with 20 imputed datasets and the default Markov Chain Monte Carlo (MCMC) imputation algorithm.
Accordingly, a simple and efficient nonparametric iterative imputation algorithm, called the NIIA method, is designed for iteratively imputing missing target values.
