On Mon, 21 Oct 2002 16:53:31 +0100, "Darren Williams" <[EMAIL PROTECTED]> wrote:
> OK firstly I don't know much about statistics and I was hoping some kind > soul here would lend a hand. > > This is to analyse a set of biological data. I'm using the Stats Direct > software package - a beginner's software app apparently. > > I have a set of data. Its is looking at the degree of expression of a marker > (cell antigen) in two different populations (EBV+ and EBV-). > > The degree of expression of the marker is divided into: > None, Weak, Moderate, Strong, Very Strong (1 - 5) > > A typical data set is: > > Marker: CD5 EBV+ EBV- > 1 2 5 > 2 4 3 > 3 6 2 > 4 12 15 > 5 9 12 > > > I want to look at two things: > > Does EBV status effect the degree of expression of CD5. > I have used Fisher's Exact test here, is that the correct test to use. [ later note says: 5x2 contingency chisquared, not Fisher's which is defined for 2x2 tables.] A 5x2 contingency chi squared test is a general test for any difference in pattern. Presumably (if not weird data), you will have better power and a more sensible statement when you compare the scores of 1 to 5 on 'strength of expression'; and conclude that the means - which you can illustrate as being almost identical - do not differ by a t-test, or by an interesting amount. -- 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/ . =================================================================
