[SciPy-user] time series analysis

Pierre GM pgmdevlist@gmail....
Tue Nov 6 14:17:07 CST 2007


Matt and I started TimeSeries each on our own, before merging our efforts last 
Christmas. On my side, I was trying to find something equivalent to pyclimate 
for my own purpose, but supporting numpy. I poked around this package a bit, 
wasn't completely happy with it, then figured it would be as easy to redo on 
top of numpy. It's while I was struggling with subclassing the masked arrays 
of numpy.core.ma that I decided to reimplement maskedarray, which eventually 
leads to the current version of timeseries.

If I'm not completely mistaken, numpy.core.ma is a translation via numeric of 
Paul Dubois' s implementation for CDAT. So yeah, I could have followed that 
path, but I was learning python at the same time and it was a very good 
exercise to reinvent the wheel, thus adding more noise and confusion.

I do agree that there could be some linkage between timeseries, pyclimate and 
CDAT, at least in terms of converting objects from package to another. 
However, it's unlikely to happen any time soon. Nevertheless, I may try to 
implement some of the functions of pyclimate and CDAT in numpy/scipy, when 
I'll have the need for them.

What I would suggest you is to start with timeseries, as it's pretty easy to 
use. Then, depending on what your needs are, you can start developing your 
own functions.

In any case, don't hesitate to contact Matt or myself if you need some 
specific help. The TimeSeriesPackage page of the scipy wiki is a good start, 
Matt did a terrific job.


More information about the SciPy-user mailing list