missingness

Missingness refers to the absence of observed values for one or more variables in a data set. It is a central concern in statistics because how missing data are handled can influence conclusions. Missing data are commonly categorized by mechanism: MCAR (missing completely at random), MAR (missing at random), and MNAR (missing not at random).

Patterns of missingness include item nonresponse (omitted answers on questions) and unit nonresponse (no data from a participant). In longitudinal studies, dropout can produce monotone missingness; other designs can yield non-monotone patterns.

Common strategies include complete case deletion, available-case analysis, and imputation. Single imputation methods (mean or regression) fill in values; model-based approaches such as maximum likelihood and multiple imputation generate several plausible datasets and combine results. MAR or MCAR underlie many standard analyses; MNAR requires explicit modeling or sensitivity analysis.

Missingness reduces statistical power and can bias estimates if related to the outcome. Diagnostics and sensitivity analyses assess robustness, and patterns of missingness are routinely reported in research. The term is widely used in statistics, epidemiology, survey research, and data science.