On Mon, 27 Feb 2006, Arnaud Bervas wrote: There have been some requests/suggestions/questions regarding this in the past. To my knowledge, nobody has done anything along these lines yet, although as Jack has said, the new matrix manipulation facility in (not-yet-released) gretl 1.5.1 makes this possible, if not exactly easy. I've recently got interested in the subject, partly because it seems to offer a systematic way of offering Exact Maximum Likelihood estimation of ARIMA models within gretl. It's possible we'll add the Kalman filter before too long, but probably not in the next (overdue!) release. Allin Cottrell