Predictions from dynamic mechanistic models used for process monitoring and control often exhibit sustained offset from process measurements. This offset is caused by imperfect measurements and by model deficiencies that result from simplifying assumptions, unmodeled disturbances, and uncertain parameter estimates. Extended Kalman filter (EKF) state estimation can eliminate offset by on-line updating of a subset of the model parameters. Offset elimination is accomplished by incorporating nonstationary stochastic states in the model equations. A difficult problem faced by practitioners when implementing state estimators is deciding which model parameters to update using on-line measurements. In this article, simple screening tools are developed to aid in updateable parameter selection. These tools are extensions of the relative gain array (RGA), the relative disturbance gain (RDG), and the disturbance condition number (DCN), which have been used in multivariable control applications to determine appropriate manipulated variable/control variable pairings and to examine disturbance effects. The application of these techniques for updateable parameter selection is demonstrated using simulations of a gas-phase polyethylene reactor system. A benefit of these screening tools over past trial-and-error parameter screening practices is that neither tuning of the state estimator nor running of simulations is required. We show that the RGA is an effective tool for determining when problems will arise due to correlated effects of different parameters on model outputs. The RDG is shown to be an effective tool for reducing the number of adjustable parameters when only particular types of disturbances are anticipated. We demonstrate that the DCN can be used to screen out parameter sets that will lead to excessive and physically unrealistic adjustment of model parameters.