Blue = observed, grey = missing
P(observed) given X

Missing Data Mechanisms: Observed and Missing Cases (2D)

Delete values from a data set by three different rules and see which ones bias the average. Missing completely at random is harmless, missing at random is repaired by regression on a helper variable, and missing not at random is not. The closed forms for probit response are checked against large simulations. A Blender-built plotting easel with data jars, numerical checks built in.

Made with 🟠 Blender