Chapter 18 Missing Data

Missing Data happens. Not always

• General: Item non-response. Individual pieces of data are missing.
  
• Unit non-response: Records have some background data on all units, but some units don’t respond to any question.
• Monotonone missing data: Variables can be ordered such that one block of variables more observed than the next.

This is a very brief, and very rough overview of identification and treatment of missing data. For more details (enough for an entire class) see Flexible Imputation of Missing Data, 2nd Ed, by Stef van Buuren: https://stefvanbuuren.name/fimd/

This section uses functions from the following additional packages: mice,MASS, VIM, and forestplot.

Some examples use a modified version of the Parental HIV data set (Codebook) that has had some missing data created for demonstration purposes.

library(VIM); library(mice)
load("data/mi_example.Rdata") #not available to public