Skip to content
Prev 19416 / 20628 Next

R-sig-mixed-models Digest, Vol 175, Issue 23

Not a complete answer, but some thoughts that other people might 
comment on.

   If you're primarily interested in treatment effects, because your 
treatments are applied at the level of fields, it might make your life a 
lot easier to compute average values by field & day (see Murtaugh 2007). 
Your data are (1) Gaussian (or being treated that way) (2) balanced.

   I can't see from your post whether the fixed effect model is additive 
(Treatment + poly(DAT, 2)) or interactive (Treatment*poly(DAT,2)).  You 
don't really need emmeans/emtrends to test the overall effect of 
treatment: you could use anova() to compare the models

NumberInsects ~ Treatment * poly(DAT, 2) + ...

with

NumberInsects ~ poly(DAT,2) + ...

   It's not so important if your primary interest is in Treatment, but 
you have at least the potential to consider variation in the temporal 
patterns across plots and subplots (random effect  (poly(DAT, 2) | 
Plot/subplot) ), if you have sufficient data ...


Murtaugh, Paul A. ?Simplicity and Complexity in Ecological Data 
Analysis.? Ecology 88, no. 1 (2007): 56?62.
On 7/27/21 1:54 PM, Inka Willms wrote: