inst/vignettes/convergence2.Rnw is the most comprehensive test suite. It assesses the
accuracy of fitted transformations versus the known true ones, using
simulated data.

Then there are some more basic tests and debugging platforms for the
ML estimators:

testmlest.R is a script that checks whether the ML estimate of the
parameters a and b, if mu_k and sigma are known, is correct, based on
simulated data. It also plots the likelihood landscape around the 
true parameters.

testderiv.R tests whether the explicit gradient of the neg
loglikelihood that is used by in the vsn code accords with the
numerical differentation of the neg loglikelihood.

testprofiling.R tests whether the ML estimate of the parameters a and
b, if mu_k and sigma are known, is correct, based on a previous
profile LL fit.
