Missing Data Mechanisms: Complete-Case Bias under MCAR, MAR and MNAR and Mean-Imputation Shrinkage
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.
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.
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.
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install