Fully-Observed Covariates

If your imputation dataset has variables for which no observations are missing, such as fixed effects or survey design variables, ${\mathfrak{A}melia}$ can more efficiently generate imputations if you specify the columns in which the completely observed variables appear using the global _AMfully. The imputation model is still multivariate normal but the computational algorithm takes into account that no imputations are necessary for these variables. This reduces the number of parameters that need to be estimated so is especially useful when the number of variables in the imputation model is high relative to the number of observations in the dataset.

If the number of partially observed variables is $p$ and fully observed covariates is $c$, then the number of parameters to be estimated in Amelia is $p(p+3)/2 + p(c+1)$.