future work: model assessment as finding the roots of a p-value function

[jwbowers]

It might speed things up if we tried to find the hypotheses for which evidence is of a particular level (p-value) without doing a grid search. The uniroot function would do this for a one-dimensional problem. So far uniroot.all (from rootSolve) and BBsolve (from BB) require that the inputs and outputs of the function be the same length because they seem themselves as solving systems of equations. As far as I can tell online, people just do a coarse graph of a function and then give uniroot different intervals (Say for the upper and lower boundaries of a CI). Optimization routines want to give only one answer. We will typically have many answers (imagine the .05 contour on a levelplot of p-values).