Errors with Patsy formulas

[edaub]

The updated GP uses patsy to create design matrices. The implementation currently only supports providing the RHS of the formula, but should be extended to use the dmatrices option in patsy to allow use of formulae such as "y ~ x[0]" to specify both sides of the mean function regression to get the appropriate design matrix.

Current workaround is to remove the LHS of the mean function formula. This is likely to only be a minor inconvenience as this is ignored in all invocations of the code.