FANTER? (adehabitat)

I have a doubt. I'd like to implement the FANTER analysis, described in
Calenge&  Basille (2008), which should be a type of Gnesfa analysis, right?
But I don't know how to implement it (in adehabitat)... the gnesfa default
option is equivalent to FANTER?

No. Actually, depending on the distribution chosen, the GNESFA will 
correspond to the MADIFA or the FANTER.  Consider the examples of the 
help page of this function:

## Loads the data
data(bauges)
kasc <- bauges$kasc
locs <- bauges$locs

## Prepares the data for the GNESFA:
litab <- kasc2df(kasc)
pc <- dudi.pca(litab$tab, scannf = FALSE)
Dp <- count.points(locs, kasc)[litab$index]


In this case, pc stores the environmental information. Conceptually, it 
can be considered as a table storing the value of the environmental 
variables (columns) in each pixel of the map (rows). Dp is a vector 
containing the utilization weights, i.e. the number of animals in each 
pixel of the map. The MADIFA corresponds to a GNESFA with the reference 
distribution corresponding to the utilization weights, that is, to 
perform the MADIFA, type:

gn <- gnesfa(pc, Reference = Dp)

If you want to perform a FANTER, you have to set the utilization weights 
as the Focus distribution, that is:

gn <- gnesfa(pc, Focus = Dp)

  I understand the modifications leading to
ENFA and MADIFA (using gnesfa fuction), but I'm kind of lost in how to
implemet FANTER...
I know (following the paper) that I should keep the first and last
eigenvalue, but what about the other options of the function?
   

You can choose the number of first and last axes that you keep in your 
analysis, not necessarily only the first and last one.
The options nfFirst and nfLast are easier to understand if you do not 
set scannf=FALSE, so that the eigenvalue barplot is displayer. For 
example, if you can identify visually a clear "break" in the decrease of 
the eigenvalues after the second eigenvalue, then, it would be a good 
idea to keep the first two axes. Similarly, if you can identify a strong 
"break" in the increase of 1/eigenvalues just before the eigenvalue P-3 
(where P is the total number of eigenvalues), then it would be a good 
idea to keep the last three axes. Then factorial maps and other tools 
described on the help page and in the paper would help to interpret the 
results.
Hope this helps,


Cl?ment Calenge