(Reply sent to OP and the list.) The term "repeated measures anova" can refer to a multitude of designs, one of which would surely be appropriate. I should think, from your description, that you'd want to do a two-factor 4-by-6 ANOVA with the 4-level factor (time, or season) as the repeated measure and the 6-level factor (colony, or depth) as the "between-subjects" factor. Since you have only one "subject" in each of the 24 cells, the interaction between the factors cannot be assessed, and the formal interaction mean square (with 15 d.f.) serves as the error mean square. If the "colony" (or "depth") effect were negligible, you might re-analyze the data as a one-way repeated measures design (omitting the between-Ss factor), to see whether the additional 5 d.f. for error produce enough greater sensitivity (despite the larger error SS) to make a difference. (If I were doing it, I'd follow this with a median polish (Mosteller & Tukey, Data analysis and regression (1977), if I remember correctly) to see whether there be evidence of interaction in the residuals, and if so I might then manually adjust the error mean square to remove the influence of one or two cells (which would of course reduce the numer of degrees of freedom for error by one or two). But that's maybe more than you wanted to know.)
If you're interested in correlations between the measured variables on the 9 chemicals, you might do this as a MANOVA with 9 (or some multiple of 9 -- you didn't say whether the response variables be merely the presence of each chemical, or its concentration, or some other measure(s) of the chemical). A MANOVA always produces the relevant univariate ANOVAs as part of the output. If the measurements are merely presence/absence of the chemical, you might want to carry this out as an ANOVA analogue of logistic regression; although it seems to me that in that case you may not really have enough data to do a decent job of modeling probabilities... Good luck! -- DFB. On Tue, 9 Mar 2004, Ron Devine Vave wrote: > A friend is doing a field study, looking at six naturally occurring > sponge colonies (all of the same species) on a small reef section > (each at different depths). In each sampling time (4 in total, to > determine seasonal effect), he took out a small piece of each sponge > and extracted chemicals from it. in total, there were about 9 > chemicals. > > Will it be okay to do a repeated measures anova with the six sponge > colonies as subjects and using time as within-subjects factor, and the > different chemicals as measurements? This is, despite the fact, that > the 6 sponge colonies were at different depths and could have had an > effect on the amount of chemical produced. ------------------------------------------------------------ Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
