[BioC] Arguments for normalizewithinarrays (Limma)

[Gordon Smyth]

Dear Sally,

Here are two pieces of advice which apply generally to most functions 
in Bioconductor or R, and certainly to limma.

Firstly, tuning arguments are usually set to default values which are 
sensible for most purposes. You would have to have a special reason 
for choosing something different. You have not mentioned any reason 
why the default settings of normalizeWithinArrays are not suitable for you.

Secondly, if you want to learn more, you read the help page for that 
function and follow the references given. In this case,

   ?normalizeWithinArrays

The 3rd reference looks especially relevant.

BTW, reading the help page would tell you that the 'df' argument is 
not used for loess normalization.

Best wishes
Gordon

>Date: Wed, 24 Oct 2007 18:28:19 -0700
>From: Sally <sagoldes at shaw.ca>
>Subject: [BioC] Arguments for normalizewithinarrays (Limma)
>To: bioconductor at stat.math.ethz.ch
>
>I am trying to understand what values I should put in for the 
>following arguments.  For example how many iterations should I 
>use.  What is a good value for span, df etc.?  Can anyone guide 
>me?  Is there a good reference source for this information?
>
>normalizeWithinArrays(object, layout, method="printtiploess", 
>weights=object$weights, span=0.3, iterations=4, controlspots=NULL, 
>df=5, robust="M", bc.method="subtract", offset=0)