I have been working with "lmer" on some mixed effects data, and need a
little advice sorting out the random effects.
I'm interested in the variations in relative abundance of two insects, one
of which is rare. So, I counted each (SpA and SpB) during nine visits to
4-12 traps at 8-12 sites, total of ~350 records.
I've considered using the ratio SpA/SpB as an independent variable, but then
over half the values will be zeros and some division by zeros. Is it
reasonable to use SpA/(SpA + SpB)?
I have used these random effects: (1 | siteID/trapID) + (1 | factor_visit).
However there is a complication. SpA likes buildings more than bushes, and
SpB is the opposite, so each trap is graded on a 1:3 scale I have called
"in_out". I started with "in_out" as a fixed effect, and it drops out as not
significant. I am pretty sure it is significant but is being absorbed in the
site/trap random effect. How to fix this?
[[alternative HTML version deleted]]