I forgot to mention that I believe this issue will be found in j1(3) as well. I haven't tried any tests with y0, y1, jn or yn. I'd be might surprised if these were robust. -- Steve http://troutmask.apl.washington.edu/~kargl/