Strictly speaking, statisticians distinguish between two forms of missing at random:

1. Completely missing at random occurs when the reason for absence of an observation is independent of both the observed measurements and the unrecorded outcome.
2. Missing at random occurs when absence of an observation depends on some covariate(s) values, but not otherwise on the unrecorded outcome.

The distinction is really only required when using the multiple imputation methods commented briefly on below.