In this work a new data compression method is presented which is applicable to systems whose observables are a linear combination of only a part of the system states. This characteristic is quite common in applied problems, especially in inertial navigation systems (INS). As a result, the formulation of the Kalman filter can be revised to yield a peculiar form for the covariance and state update which is the foundation of the new data compression method. The computations, according to this method, are divided into fast and slow rates. At the fast rate, which is determined by the availability of the measured data, a reduced-order Kalman filter is propagated and updated. The full-order system is propagated and updated only at a slow rate chosen by the designer. Utilizing the peculiar Kalman filter form of this case, the full-order system update is performed on the basis of the output of the reduced-order filter at each slow rate update time. Results of the application of this new data compression method to an INS are presented.