The covariance projection equations (error budgets), which express the
error covariance as the sum of covariances due to different error sou
rces, are derived from the sensitivity analysis equations of optimal e
stimators for batch processors, filters, and smoothers for discrete an
d continuous problems, The sensitivity analysis of optimal estimation
covariances are shown to be the same as previously derived suboptimal
estimation sensitivities, which characterize the loss of estimator per
formance due to the use of nonoptimal values for error covariances. Pr
ojection equations for steady-state filter and smoother values are obt
ained, Some steady-state filter examples are considered.