simplify the model equation code a bit

[drbenvincent]

At the moment we have some silliness in the model equations because we are extracting information from pandas DataFrames. Ideally we'd be able to use dot indexing (ie attribute look up) so you can do

θ.logk

rather than

θ['logk'].values

The same thing applies to the designs which are also held in a pandas DataFrame. For example, we have to do this data['DA'].values rather than data.DA.

This is just unnecessary faff which makes the models less clear than they should be.

Ideas

1. One idea is that I'm not using Pandas properly.
2. Use numpy record arrays because they allow for dot indexing.