emchristensen/plotswitch

Analyzing the effects of the 2015 Portal plot treatment switch on the rodent community.

plotswitch's Issues

check final GAMs

I want to have one more check on the final GAM models we're using to make sure they're appropriate for what we're trying to do.

notes

community-level measures vs community composition
look at murids on new controls
solve autocorrelation problems in GAMs
error structure: neg binomial?

meeting 5/18

1a. run models using data at switch and later

1. combine AR and Poisson approaches
2. keep separate file for GAMM (clean and document)
3. pull repeated things out into functions, put in gam_functions.R
4. current AR model is within year AR; is there across-year autocorrelation?
5. investigate link functions (why log with poisson)
6. figure out how to compare models using different families
7. what is the difference between gam and gamm? do they produce different results with same arguments?

plot-level or aggregate

we need to decide if we want to model the site as an aggregate (which may have some issues), or use plot-level data (which means plot needs to be included as a factor in the GAM)

do different plot treatments have different plant communities?

Need some sort of analysis (pcca?) of whether plants still differ by rodent treatment just before the switch (i.e., could the indirect effect of rodents changing plant community contribute to the reduced ability of dipodomys to re-invade plots)