I have carried out an experiment that measured soil chemistry parameters at different times (0, 3, 6 and 9 months) from 12 plots in the field (3 treatments arranged in a fully randomized design with four replications). In each case, soil was randomly sampled in ten squares to give a composite soil sample per plot. Therefore, I believe that each soil is different. I statistically analysed with two-way ANOVA (with treatment and sampling date as factors). However, a colleague told me that since sampling dates are not independent from each other, it doesn’t make sense to use a two-way ANOVA with treatment and sampling date as orthogonal factors. Instead, a repeated measure ANOVA should be used. Which is correct?
Repeated measures ANOVA is more appropriate. Your data are certainly correlated, since they are the same plots over time. Consider as well using a general linear mixed model or a covariance structure model as well, depending on whether the assumption of compound symmetry (used in RM-ANOVA) is valid. Good luck!