The VIM
package allowed us to visualize the missing data patterns. A related term, the missing data mechanism, describes the process that determines each data point's likelihood of being missing. There are three main categories of missing data mechanisms: Missing Completely At Random (MCAR), Missing At Random (MAR), and Missing Not At Random (MNAR). Discrimination based on missing data mechanism is crucial, since it informs us about the options for handling the missingness.
The first mechanism, MCAR, occurs when data's missingness is unrelated to the data. This would occur, for example, if rows were deleted from a database at random, or if a gust of wind took a random sample of a surveyor's survey forms off into the horizon. The mechanism that governs the missingness of drat
, mpg
, cyl
, wt
, and vs
is MCAR, because we randomly selected elements to go missing. This mechanism, while being the easiest to work with, is seldom tenable in practice.