Thanks for your answer.
The package TimeSeries was designed to perform the kind of operations you want (filtering, frequency conversion, handling missing data...). It is currently available in the sandbox of scipy. More info here: http://www.scipy.org/SciPyPackages/TimeSeries Wow, this looks great. But a little complex ;-) Well, one could write functions for common tasks that fascilitate it a bit...
* filling data gaps for days or hours without valid data using statistical methods like regression
This one might be a tad trickier. You may want to give pyloess a try, it's a wrapper around the dloess functions. Also available in the scipy sandbox.
Any idea when there will be a first binary release of all this together with maskedarray? Has anyone else done this kind of things? Maybe with a interface to R? Regards, Timmie