Hello!
I've been using the agricolae package for ANOVA and LSD tests for data from split-plot designs with randomized complete block design of main-plots. I would greatly appreciate any advice on how to analyze simple effects when there's a significant main-plot x sub-plot interaction. In the R tutorial (
https://cran.r-project.org/web/packages/agricolae/vignettes/tutorial.pdf) on p 36-37, a multiple mean comparisons with a factorial treatment structure is performed for a model with a significant interaction (clone x nitrogen):
outAOV <-aov (yield ~ block + clone * nitrogen, data = A)
# this does LSD on the "simple effects" as opposed to "main effects" of clone and nitrogen separately:
outFactorial <-LSD.test (outAOV, c("clone", "nitrogen"), + main = "Yield ~ block + nitrogen + clone + clone:nitrogen",console=TRUE)
My questions: 1. Does this also apply to split-plot designs with RCBD mainplots? (see p 8
for a visual example of this general layout, https://stat.ethz.ch/education/semesters/as2011/anova/Slid10.pdf) 2. Do I have the correct error term specified for the mainplot in Error() below, or is it supposed to be Error(block:mainplot)?
# General model structure for split-plot with RCBD mainplots
outAOV <-aov (yield ~ block + mainplot * subplot + Error(mainplot:block), data = my.data)
# When significant interaction, analyze simple effects
# However, the following produces an error message, and I think it's because of the error term. I could take it out, but I'm not sure that's correct
outSplitPlot <-LSD.test (outAOV, c("mainplot", "subplot"), console=TRUE)
I could also subset the data for each level of mainplot and analyze the subplot effects (and vice versa), but that reduces the power of the test.
Thanks!
-Maegen
--