printtiploess

The print tips may become deformed during the printing process. Within-print tip normalisation corrects for this effect by fitting a loess curve on each group of spots in the array printed specifically by that print-tip. The normalised values are given by:

$\displaystyle M_i=M_i-$loess$\displaystyle _i(A),$    

where $ i, \;\;\; i=1,\cdots,I$, donates the number of the print tip group [7].

According to [4], one should not use print-tip normalisation on Agilent arrays, where one should use global loess instead. This method is also unreliable for small arrays with less than $ 150$ spots per print-tip group. Larger arrays may behave like this if the number of spots with non-missing M-values is small for one or more of the print-tip groups. If this is the case, one should use global loess, or the empirical Bayes compromise between print-tip and global loess normalisation, ``robustspline''.



Bjørn Kåre Alsberg 2006-04-06