Hello,
I have a sample design that I'm having trouble translating into the correct syntax in the mgcv package language. My set-up looks like this: We trawl for fish in a bay to see if they're eating or not (empty or not-empty: 1/0) and see if it's linked to the location and/or time of capture. We trawl once per "station" (the location), at 80 stations, once per month, every year. Several stations make-up a zone. If I wanted to know if fish with empty stomachs varied between zones and years, but not stations, would that look something like this? Can a variable be both a fixed and random effect?
library(mgcv)
# fStation = factor location
# fZone = factor zone
# fCYR = factor calendar year
# CYR = year as continuous variable for random slope
bam(empty ~ fZone +
fCYR +
s(fStation, fZone, fCYR, bs = 're') + # station within zone, within year?
s(fZone, CYR, bs = 're'), # Each zone can have its own trend through time?
data = c_neb4,
method = 'fREML',
discrete = TRUE,
family = binomial(link = "logit"),
select = TRUE)