[BioC] Limma 1.6.1 : composite normalisation method not recommended ?? (+ function error)

[Laetitia Marisa]

Hello,

I am trying to normalize my data from titration controls with the 
composite method.

First of  all I would like to know why this method is not recommended 
any more by the author (in a previous mail from Gordon Smyth).

Otherwise here is my problem if it is of any interest to go on with this 
method : the function returns an error that has already been mentioned 
in the mailing list but it seems to me that I don't have a problem with 
my controlspots variable.


Here is the command line I am using :

nMA <- gif.normalizeWithinArrays(MA, layout=MA$printer, 
method="composite", controlspots=ctrlspots)

where ctrlspots :
str(ctrlspots)
 logi [1:864]  TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE  TRUE  
TRUE FALSE FALSE ...

and MA$M  :
str(MA$M)
 num [1:864, 1:6] -0.2679 -0.1382  1.3026  0.0359 -0.9560 ...


The error returned is :

Number of Observations: 192
Equivalent Number of Parameters: 4.95
Residual Scale Estimate: 0.12
Error in predLoess(object$y, object$x, newx, object$s, object$weights,  :
        NA/NaN/Inf in foreign function call (arg 3)

This error comes from the predict function : predict(fit, newdata=x)
There are infinite values in x in my data and in the function there no 
treatment with.

Thanks a lot,

Laetitia.