When to use missing data imputation in the data analysis problem?

Question

I want to run statistical analysis of a dataset and build a logistic regression model and multinominal linear model by R according to the research question. But I was wondering which step should I use the missing value imputation to complete the dataset. I have finished the univariate analysis for each variable in the raw dataset, and I found there are three continuous variables and two categorical variables with lots of missing data. I want to use the missing data imputation to complete the dataset after processing with the bivariate analysis and graph exploration of every variable. But I am not sure if that's a correct order to do?

Should I use missing value imputation to complete the dataset before bivariate association analysis or should I do it after that?

In addition, if I want to examine the distribution of the outcome variables to find proper transformation, should I do it after impute the missing data as well?

Thank you!

Answer 1

Generally speaking you have two options:

• impute the missing data
• discard the missing data

Due to the fact that ML models perform better with more data, the former is usually preferred. However, you should keep in mind that the imputed data should not affect the distribution of the feature.

This is especially the case with features with a high percentage of missing values. If a feature has, let's say, 90% of its values missing, then by imputing it you are dictating what you want its distribution to be like (because all the data in that feature will be artificial). In this case it is better to discard the feature altogether.

Answer 2

Imputing missing data (that is, filling in missing values with some other value) is not appropriate for analysis or regression. It would only be valuable if you were going to try to train a learning model for predictions. Putting in random or inferred values would mess with your analysis