I have a 4898*17002 variables dataset. I subset as follows:
Dataset 1(4898*70): All variables have missing values in the form of NA (Generally in the range of 30-50%). I use mean and median imputation techniques however I don't want to use these techniques as they will introduce a bias.
Dataset 2(4898*10): Subset of dataset 1 (Attention, Self-control, Anxiety, and Depression have been constructed from the variables in dataset 1 and are as follows:
Gender Cognitive_Score Attention Grit Self-control Anxiety Depression Income Education GPA
My question is that how do I use multiple imputations? Should I use it in dataset 1 before adding up the variable scores to form my constructs (Self-control, Grit, etc.)? (Note: all the variables have at least 30% missing values as some people did not take the survey in some of these years whilst others were missing. All of them have been clubbed into NA). I will be using Amelia package in R for imputation.