We imputed missing BMI values using linear regression on age and sex for complete cases, and the missing values for other categorical variables were imputed by taking the mode values after adjusting for age and sex.
The incomplete data set is completed by iterative imputation of the missing values with the corresponding imputation model.
For the RDQ, QOLRAD, and SF-36, where at least 50% of items in a dimension were completed, the mean value of the completed items was used to impute the missing values.
To handle the missing values, we evaluated the complete-analysis adjustment method and all mean-based imputation algorithms described by Luengo et al.[ 38].
Because missing values can hinder downstream analysis steps that require complete data as input, it is crucial to be able to estimate the missing values.
Namely, for each dataset, we randomly removed 5% entries of the complete matrix to generate a matrix with missing values, and then estimated the missing values using the shrinkage and the original regression-based methods.
Namely, for each of the six benchmark dataset, we randomly removed x% (x = 1, 5, 10, 15 or 20) entries of the complete matrix to generate a matrix with missing values, and then estimated the missing values using the shrinkage and the original regression-based methods.
This model is used to generate several replicate 'completed' data sets, containing imputed values in place of the missing values.
M-Step: We replace the missing value by in the complete log-likelihood function Eq. 7. Then we maximize the function: 12 Regular approaches can be used to obtain the maximum likelihood estimator of parameters in Eq. 9. Starting with proper initial estimates of the parameters, we iterate between E-step and M-step until convergence is achieved.
