I think we need a rather more concrete description of the problem than you've so far provided, in order to offer any sort of useful advice.
On 16 Oct 2002, Jose H. Vos wrote: > At the moment I'm analyzing a number of datasets You wish to analyze them one at a time, or did you intend to deal with them simultaneously (or, perhaps, both)? > containing repetitive measures of ratios. "Repetitive" in what sense? Do you mean to invoke the idea of "repeated measures" in ANOVA designs? Ratios of what? (Of frequencies, e.g.; or of mass to volume; or of length to width; or ... ?) > We have identified 7 classes of cell stages in animal tissue. Do these classes constitute something like a design factor, or are they categories that might be observed, and therefore represent a categorical dependent variable? Or have I misunderstood wholly? > The animals were treated with different concentrations > of hormone, each concentration replicated 3-4 times. Same set of concentrations for all animals, or different concentrations for different animals, or what? And presumably the order of treatments, and the time between treatments, is of some importance? > I have tried the G-square test Sorry, I don't know what this is. Has it a different name; is there a standard reference; and/or can you describe it briefly? > but in this test I could only work with average cell ratios per > treatment and I would really like to insert all variation in the > dataset. It rather sounds as though some variant of ANOVA would be desired. > Other problem is that we have no "expected" cell ratio. Now I used the > control cell ratio as expected cell ratio for the F0. I understand about "expected frequencies" in the context of (say) a chi-square test of independence of classification systems. I don't understand "expected ratio", which may simple reflect my lack of familiarity with the G-square test. > Has anyone experience with similar datasets or an idea how to test > these data described above? Is the G-square test an appropriate test > for the ratio data? >From my perspective, there is not enough _description_ above to know what a "similar dataset" would look like, let alone what might be a suitable analytical scheme. By the way, and perhaps in the service of pedantry, one does not test data. The question you (apparently) seek to address is how to test an hypothesis; and since you haven't specified the hypothesis (-es?) you want to test, it's a trifle difficult to address the question. ----------------------------------------------------------------------- Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 [Old address: 184 Nashua Road, Bedford, NH 03110 (603) 471-7128] . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
