I'm a frequent R user and ran into a message from 2004 in which someone
wants to use R for a Siegel-Tukey test. To the best of my knowledge (and
I've checked a lot), there is no way to do this in R. However, I've set up
some R code that requires the sue of an executable program made with a
general-purpose language (turing). Ultimately, the test is performed
correctly a la Conover and/or Sprent. I will happily make the whole thing
available to anyone who wants it.
I have a question. Suppose the two data sets have different means and the
difference is unknown a priori. Can the difference be estimated from the
data when using the Siegel-Tukey test? Reputable sources disagree on this.
What do you think? I say yes. It isn't as if we never use sample standard