[SciPy-User] scikits.timeseries question

[Christopher Barker]

Folks,

Sorry ask a question and then disappear, but I had a full day away from 
work and email yesterday! I do appreciate all the help, though. It's 
going to ake a bit to digest this, but it looks like I can do what I 
want to do here.

Pierre GM wrote:

>>> np.array(...) + sd gives you a ndarray of Date objects (so its dtype
>>> is np.object), and you use that as the input of date_array. The
>>> frequency should be recognized properly.
>> OK -- though it seems I SHOULD be able to go straight to an DateArray, 
>> and I'm still confused about what this means:
> 

> Well, that depends on the type of starting date, actually. If it's a
> Date, adding a ndarray to it will give you a  ndarray of Date
> objects. If it's a DateArray of length 1, it'll give you a DateArray.

ah -- that does look helpful -- I guess I find it surprising that there 
is a difference -- I think this may be legacy issues -- what I imagine 
we'd want is a DateArray, and there would be a corresponding DateScalar 
-- like there is for an array of float128s or whatever.


>>>        * a sequence of integers corresponding to the representation of 
>>>          :class:`Date` objects.
>> That's what I have: a sequence of integers corresponding to the 
>> representation of the Date objects (doesn't it represent them as "units 
>> since start date" where units is the "freq" ?
> 
> No, not exactly: the representation of a Date objects is relative to an absolute build-in reference (Day #1 being 01/01/01). (Likewise,  nump.datetime64 uses the standard 1970/01/01). 

OK -- got that now. I suppose I could add the integer value of my start 
date and then I"d get what I want.

>> Definition:
>>        ts_lib.mov_average(data, span, dtype=None)
>> Docstring:
>>     Calculates the moving average of a series.
>>
>> Parameters
>>     ----------
>>     data : array-like
>>         Input data, as a sequence or (subclass of) ndarray.
>>         Masked arrays and TimeSeries objects are also accepted.
>>         The input array should be 1D or 2D at most.
>>         If the input array is 2D, the function is applied on each
>>         column.
>>
>> I've got a 3-d array -- darn! Maybe I'll poke into it and see if it can 
>> be generalized.
> 
> 
> 3D ? What are your actual variables? Keep in mind that when we talk about dimensions with time series, we zap the time one, so if you have a series of maps, your array is only 2D in our terminology. 
> If you have a time series of (lat, lon), mov_average will average your lats independently of your lons

Then is should work -- what I've got is a 3-d array in a netcdf file, it 
has dimensions: (time, lat, long) -- so it sounds like that should work.

Robert Ferrell wrote:

> Is the issue that sd is a Date and not a DateArray?  You can always  
> make a DataArray with sd, of the correct length, and then add to that:
> 
> In [83]: sd = ts.Date('d', '2001-01-01')

> In [86]: d1 + array([0,2,3,5])
> Out[86]:
> DateArray([01-Jan-2001, 03-Jan-2001, 04-Jan-2001, 06-Jan-2001],
>            freq='D')

yup - that's what I want! Thanks.

> In any case, I get the impression that the OP has an array of integer  
> offsets generated in some other fashion entirely.

exactly.

Matt Knox wrote:
> In addition to the methods described by Pierre and Robert, you could also do:
>     
>>>> sd = ts.now('d')
>>>> relative_days = np.array([1,5,8])
>>>> absolute_days = relative_days + sd.value
>>>> darray = ts.date_array(absolute_days, freq = sd.freq)
> 
> which I think probably has the lowest overhead (but don't hold me to that :) )
> if that matters for your application.

Frankly, I don't think it does matter, but I like efficiency all the same.

Thanks to all three of you,

-Chris




-- 
Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chris.Barker at noaa.gov