Time series LME with nested fixed effects

Hi Austen,

apart from the residual error I could imagine that there are two additional layers of
variation in the experimental setup you've:

(1|seal) - variation between seals

(1|seal:season) - season-specific seal variation

(1|seal:season:period) - season- and period-specific seal variation


Far from sure that all sources will show up for an experiment only involving 5 seals. For
clarity, I would prefer using these three terms to using "(1|seal/season/period)", even
though I think the end result is this same as the latter is equivalent to:

(1|seal) + (1|seal:season) + (1|seal:season:period)


If there are no replicates (you didn't tell us about replicates...) the last term will be
redundant as it corresponds to the residual error term, which is anyway part of the model.
So in short your original model specification looks ok to me.


Christian