Hi everyone,
I have a grazing experiment which includes:
- Two topography(flat vs. slope)
- In each topography, there are three management type(grazing,mowing,mixed use)
- In each management type, there are seven grazing intensity(except for mowing)
- In each grazing intensity, there are two kinds of patches(lightly grazed and heavily grazed), mowing has no patch
- In each patch type, we randomly selected five quadrats
- The same experiments were done in two continuous years,and the position of quadrats between these two years may not be same
- We calculate community dissimilarity for each kind of patch, and we want to see how do the year, topography, management type, grazing intensity and patch type affect community dissimilarity
The question is I'm not sure whether I need to use mixed model, if I should use this method, how should I write the formula. I have wrote one:
model=lme(dissimilarity ~ year*topography*management*intensity, 1+year*topography*management*intensity|patch, data=mydata)
According to a book written by Alain Zuur, he suggest to consider all fixed effects and then select best fitted one by comparing AIC value.
I'm not very sure about random part.
Any suggestions and/or examples of how to write script for my experiment would be greatly appreciated.
The first step is to define the hierarchical structure of your experimental design. You stated this hierarchical structure very well in your bulleted list, and I encourage you to reexamine the Zuur et al. to understand how to write the hierarchical structure. Generally (very generally), my model framework is this:
Additional step (somewhere between steps 6-7 is to mumble to yourself, have models fail, repeats steps 1-6, repeat again, and mumble more.
Some may argue (rightfully so) that inclusion of a two-level factor in your hierarchical structure is not good, but I would argue that it's required due to your design.
There are lots of tutorials online for this kind of thing. In many cases you can even find tutorials specific to your field. For example, since my field is psycholinguistics, I generally recommend people Bodo Winter's and Harald Baayen's mixed effect tutorials because the examples in these are the kinds of experiments I do. If you do some Googling, it's likely that you can find tutorials specific to your own research field.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package