For my dissertation, I monitored a population of blue-tits looking at whether different pollutants and environmental factors had an effect on breeding success, and I want to know which effects are the strongest. My dataset is relatively large, with close to 100 nest boxes across 9 different sites. Is it possible to use a GLMM in this situation where I only have what I think are random effects?
Thanks
(Using SPSS)
is the breeding success a continuous variable [no of breeds] or a dichotomous variable [ie breed or not]
The way you've written your goals, it sounds like you don't want to treat pollutants and environmental factors as random factors, but rather as fixed factors.
Hi Deborah J Hilton, breeding success was measured in the number of chicks which reached day fourteen from the 2022 breeding season (so just one breeding attempt per nest) - in some nests, this was zero and in others, all (10) eggs were raised to fledge. This would be a continuous variable?
Hi Sal Mangiafico, to be honest, I'm not sure - each nest box has an individual measure of noise and light pollution and max and min temperature once the chicks have hatched. The values vary significantly between each nest box, making me think it was a random variable.
If you have multiple observations per nest box, you may want to treat nest box as a random effect. But I suspect you want to treat noise, light pollution, and temperature as fixed effects. But you may not want to include all of them in your model.
I would think this is a count of success ie as you say 10 or 0 as examples, a count of number of chicks and how this differed between nests. It is a count of number of chicks as success, as opposed to a dichotomous measure of success whereby you would have a cut-point defining success or not ie an arbitrarily chosen value may define success ie 5 chicks as an example.
I just want to make it clear that all variables we are talking about are random variables. Some variables are considered random effects, and some are fixed when modelling.
Deborah J Hilton, yes, I don't have a clear 'yes' or 'no' on whether each nest was a success. It's a difficult one, though, as the number of eggs produced by each nest also varied from 4 to 11, which I would have to consider, maybe a percentage yield and measure of success based on that.
Thank you for your help!
There will nearly always be one fixed effect..... the overall mean.
If all effects other than the mean are random, it is still a valid mixed model, so yes
Be aware of the definition of a random effect....it has to be a random sample from a population. If the levels of an effect are values chosen by you, it is not a random effect.
Also dont confuse effects with quantitative levels, ie regression effects, with random effects.
- Badges
- Science topic