Bug fixes
- Supplying case weights to tests for marginal homogeneity or
symmetry hypotheses appeared to work, but returned incorrect
test statistics and p-values; it now results in an error.
User-visible changes
- The shift algorithm is now computationally more efficient and
allows for user interrupt.
- Documentation updates.
Other changes
- The Fligner-Killeen test, the Conover-Iman test and the Quade
test performed transformations of the response variable prior
to applying the influence function; these transformations are
now done within the influence function.
