Maybe the Numpy library can be very useful within Python.See also the pandas really strong library.

However I don't really understand what kind of library you want to build exactly? Just only for movement and rumen pH data seems not really useful? My advise, do not try to reinventing the wheel, there are so much variables which needs to be taken into account when setting up a really good performing big data library.

https://pypi.python.org/pypi/numpy

http://pandas.pydata.org/

+1 for using numpy and pandas. For a classroom setting, I would recommend using numpy and pandas in an interactive iPython Notebook environment, and taking advantage of powerful built-in visualization functions via matplotlib / pylab. The seaborn package makes matplotlib output look way better, and is a relatively transparent overlay on the existing environement. I use this combination to interactively explore and analyze large datasets, and find the numpy-pandas-iPythonNotebook-pylab-matplotlib-seaborn combination creates an analytical environment that stikes a good balance between analytical power and flexibility.

I use this setup to analyze data locally on my notebook, but also to analyse data on several remote servers. You can set up any server as an iPython Notebook server, map the connection to an outgoing port, route a secure feed from the server to your notebook via an SSH tunnel, and map the feed to a local HTTP port. Once this is set up, you just point your local browser to localhost:XXXX (where XXXX is the port number to which you mapped the secure connection to your remote iPython Notebook server), and hack away. I tend to run large simulations on the remote servers (which would be impossible to run locally), and this setup lets me analyse the remote output data directly, without having to worry about coping the files around all the time.

This is also a good setup for interactively (and iteratively) designing publication-quality figures. Especially for complex figures synthesizing large amounts of data into a grid of subfigures, this setup allows you to re-generate your entire figure in seconds when new (or updated) data become available. If the figure were generated using a spreadsheet (for example), the cost of re-generating such figures with updated data would potentially be much higher.

All this stuff is very well documented on the web, so getting up and running is relatively easy.

Good luck!

http://ipython.org/ipython-doc/dev/notebook/

http://matplotlib.org/

http://wiki.scipy.org/PyLab 

Thanks for that !  I am writing next weeks exercise which will be aggregating data by time stamp us NumPy, Pandas & Matplotlib.

Is that Laval Quebec ?

I have not actually been involved (yet) in such project, our applications for funding has been turned down. Anyway, I would use R. Mind you, Python is far better language in aspects compared to R (e.g., strong typing), but the overwhelming amount of packages available for R, the live community etc. outweighs the pains of a language with weak typing (yes, of course, the most recent approach to object-orientation offers a high degree of type safety in R, but I still have not found any package that uses it). 

I got the following reference concerning packages for analysis: http://www.kdnuggets.com/2014/08/four-main-languages-analytics-data-mining-data-science.html  Python is among the top 4.

The reasons for using Python are the ease of getting agricultural students to use it, the integration with the Raspberry Pi to get data from sensors.  The worst aspect is the poor documentation of attribites for the amazing number of functions in the library. Most of our data is Time Series based and I love functions such as concat but wish I knew what it was doing so I can debug on failure.

To get closer to your goal to get standardized analysis code that researchers can share and improve, I would suggest to start a project at code sharing sites like github or sourceforce and invite other researchers in the field to use your code. Ease of access for everyone is essential.