Notes on Chapter 9
Back to notes on chapter 8
rough notes, need to be polished
- p. 299: cf. linear vs exponential survival models (goby data), Beverton-Holt vs. survival analysis
- p. 300: note * vs : for denoting interactions; notations differ, but R uses a*b to signify "main effects of a and b and their interaction" and a:b to signify just the interaction (that is, a*b is equivalent to a+b+a:b).
- p. 301: figure 9.2 was uglified by PUP!
- p. 302: there's a bit of a problem with adding an interaction between two continuous covariates (x1*x2) without adding the quadratic terms (I(x1^2)+I(x2^2) at the same time, because the model isn't entirely consistent.
- p. 303: note that linear regressions and ANOVA give you different kinds of summary statistics of the same model — e.g. a table of F statistics for the effects of each variable vs. a table of Wald statistics for each parameter.
- p. 304: again note f1*f2 is equivalent to f1+f2+f1:f2
- p. 309: cf SAS PROC GENMOD and PROC LOGISTIC
- p. 310: I can deal much more thoroughly with fitting type II functional responses via GLM (i.e., using an inverse link), and with the hyperbolic vs exponential function question. See this discussion.
- p. 311: not sure "quasilikelihood" is dealt with with entirely consistently but I may not be able to deal with it sufficiently thoroughly here anyway.
- p. 312 "ratio of deviances" means the deviance difference due to a particular factor divided by the residual deviance (which is an estimate of the overdispersion factor)
- p. 313 I think AICtab can now compute quasi-AIC. I'm still uncomfortable about QAIC, although Richards 2008 (Dealing with overdispersed count data in applied ecology, J. Appl. Ecology 45:218-227, doi:10.1111/j.1365-2664.2007.01377.x) suggests it's really OK.
On to notes on chapter 10
page revision: 3, last edited: 24 Sep 2008 14:17