missings

Missings is a term used in data handling to refer to missing data values in a dataset. A missings value indicates that for a particular observation, the value for a given variable was not recorded, is unavailable, or is not applicable. Missings are distinct from observed zeros, blanks, or values that have been suppressed for privacy or other reasons.

Missings can be characterized by the mechanism that produced them and by the pattern they form in the data. Common mechanisms include missing completely at random (MCAR), where the probability of a value being missing is independent of any data; missing at random (MAR), where the probability of missingness depends on observed data but not on unobserved data; and missing not at random (MNAR), where the missingness depends on unobserved data. Patterns include unit missingness (an entire case is missing), item missingness (specific fields are missing), and monotone missingness (the absence of one variable implies missingness of subsequent related variables).

Missings arise from multiple causes, including nonresponse in surveys, data entry or transmission errors, instrument design limitations, skip patterns in questionnaires, and data integration from heterogeneous sources. Their presence can affect the validity of analyses, leading to biased estimates or distorted uncertainty if not handled appropriately.

Common approaches to address missings include data deletion methods (listwise or pairwise deletion), single imputation (substituting a single value such as the mean, median, or a regression-based estimate), and more sophisticated model-based techniques. Model-based approaches include maximum likelihood methods under MAR assumptions and multiple imputation, which draws several plausible values to reflect uncertainty. Modern practices also employ machine learning–driven imputation methods that incorporate complex relationships in the data. Transparent reporting of missingness and the chosen handling method is essential in data analysis.