[R-sig-eco] Calculating an AIC for the Fisher's log series.....

[Jari Oksanen]

On 5/10/11 20:42 PM, "Cory Redman" <corymredman at gmail.com> wrote:

> I am trying to calculate an AIC for my Fisher's log series (fisherfit).  In
> order to do this I need the log likelihood.  Does anyone know how to
> calculate the log likelihood for a specific alpha value.  I know that the
> second derivative of the log-likelihood is used to calculate the standard
> error of alpha, but does anyone know how to retrieve these values (Oksanen,
> Vegan: ecological diversity).
> 
Cory,

Peter Solymos already told you how to write the logLik.fisherfit() function.
However, I think it is difficult to find a legitimate use for this function:
what would you do with the AIC? Log-likelihood and AIC defined only up to an
additive constant, and those constants do vary. You need a model where
log-likelihood is defined in with the same additive constant. The function
used in fisherfit() ignores constant terms of type -log(2 * pi) since these
do not influence the point estimates or estimates of their standard errors,
but these would influence the level of AIC. What ever you do, don't use the
AIC to compare your Fisher models to models fit with different addtivie
constants.

Cheers, Jari Oksanen

-- 
Jari Oksanen, Dept Biology, Univ Oulu, 90014 Finland
jari.oksanen at oulu.fi, Ph. +358 400 408593, http://cc.oulu.fi/~jarioksa