search

dpritchard / frair

Tools for Functional Response Analysis
8 stars 3 forks source link

Continuous responses #27

Open dpritchard opened 8 years ago

dpritchard commented 8 years ago

This is posted here for posterity (and possible future resolution):

I like your idea of applying these curves to other systems, however… (and here is the catch) - we use the following general form in our log likelihood functions:

prop.exp = rogersII(X, a, h, T)/X … return(-sum(dbinom(Y, prob = prop.exp, size = X, log = TRUE)))

Therefore if Y is not an integer then fitting these curves will fail… Implicitly X needs to have the same units as Y (as it needs to represent the maximal Y response at that value of X): So it too needs to be an integer.

I’ve also implemented checks to ensure that the nil functions do not return predictions where prop.exp is > 1 (i.e. more prey than were supplied were eaten).

So while in principle I support fitting curvets to things like biomass (also, nutrient concentration) that is not possible with these curves within FRAIR