Poul-Henning Kamp wrote: >Have you tried giving that task to a Kalman filter ?
I haven't tried implementing it at all yet, I've only theorised about it. But anyway: how do you imagine applying a Kalman filter? It would normally be used where you have a substantial series of measurements with similar quality attributes. In the case of projecting UT1, the `measurements' available appear to be the extrapolations for today that were published by IERS in successive weeks. We'd expect the extrapolations that were published later to be qualitatively more reliable than earlier ones, and it would be strange for this purpose to look at more than two of the data points (last week's and this week's). Are you thinking of applying it to some other variable? -zefram _______________________________________________ LEAPSECS mailing list [email protected] http://six.pairlist.net/mailman/listinfo/leapsecs
