On 23 Apr 2004 15:51:38 -0700, [EMAIL PROTECTED] (Yossi Berlow) wrote: > Greetings sci.stat.edu, > > Forgive my ignorance, as statistics are not my specialty, but recently > I did an analysis of a survey that included Yes/No answers. Most of > the analysis used total yes answers per subpart and the totals were > ranked and Mann Whitney Rank Sum. > > However, I also analysed the individual yes/no questions with Mann > Whitney Rank Sum, with adjustment for a tie. > > My boss insisted that I use Chi Squared for these individual tests and > so I redid the analysis and came up with the exact same P values > consistently. > > I reasoned that nominal data and ordinal data, become the same thing > when you only have two categories or two ranks. > > Does someone know if these two tests are equivalent when dealing with > binomial data?
Yes, they are, if the Rank test is making reasonable adjustment for ties. As I recall, the textbook discussions typically point out that 'this test reduces to that test' in order to show that there is nothing else going on. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
