With two-way ANOVA, why doesn't Prism offer a nonparametric alternative? Test for normality? Test for homogeneity of variances? Test for outliers?

With two-way ANOVA, Prism offers no nonparametric alternative and does not test for normality, homogeneity of variances, or for the presence of outliers. 

There doesn't seem to be any reasonable nonparametric alternative to two-way ANOVA. The only one I can find reference to is listed below, but it is said to have very low power (1).

Two comments about alternatives:

• I suspect lognormal distributions explain many cases where the cells do not have uniform variances, or where the values are not normal or seem to contain outliers. If this is the case, the solution is to transform all the data to their logarithms, and then run two-way ANOVA on those logs. 
• ANOVA is used too often, in my opinion, to compare dose-response curves or time courses. ANOVA treats different times, different concentrations, and different doses exactly the same as it treats different drugs or different genotypes. ANOVA pays no attention to the order of the data sets. It pays no attention to trends. In some cases, when you only have a few doses or time points, ANOVA may be the best analysis. But if you have plenty of doses or time points, consider fitting a model using nonlinear regression. The results from regression, in may cases, do a much better job of answering your scientific questions than do the results of ANOVA.

1. James Scheirer, William S. Ray, Nathan Hare . The Analysis of Ranked Data Derived from Completely Randomized Factorial Designs In: . Biometrics 32 (2) / 1976. International Biometric Society, S. 429-434, ISSN 0006-341X