The argument names for response functions are getting silly...

[gavinsimpson]

At the moment we have arg1 and arg2 for parameter arg along gradients 1 and 2 respectively. This is just not user friendly!

Instead, rewrite the response distribution functions Beta() and Gaussian() so that they have definition:

function(x, y = NULL, px, py = NULL)

where x and y are the first and second gradients respectively and px and py are each a list containing parameters for species on the first or second gradients. In the case of Gaussian() we would have names

• opt
• tol
• h

and for Beta() we'd have names

• A0 (or A - haven't quite decided how closely to follow notation in Minchin (1987) )
• m
• r
• alpha
• gamma

This way the parameters map to arguments without needing to have an indicator as to which gradient each pertains in the name.

[gavinsimpson]

Commit 3992a6c does the hard work of implementing this although things fail nastily in some cases if the parameter lists are not supplied with correct names. By nastily I mean the function runs but you get NAs, sometimes silently. Bad!

Hence before I close this out checks on argument names in Gaussian() and Beta() are needed.

[gavinsimpson]

7b4619f implements input checking in Gaussian() and Beta() and 16dfd1a implements a suite a unit tests to ensure failure with incorrect input.

I assume these unit tests won't cover every possible issue but they are good enough for now to close this issue.