Hello, I've been performing cyclic lowess normalization (Bolstad '03) on my experimental group data. When carrying out pairwise lowess normalizations "NA" values affect the outcome, i.e. they become additive. Has anybody any solution for this problem?. One solution could be replace NA with its replicate (biological/technical) in another array, at least for the fit computation process? Toni Antonio Juli? Unitat de Recerca de Reumatologia Institut de Recerca Vall Hebron Pg Vall Hebron 119-129 Barcelona, Spain Tel: 93 274 6200 e-mail: ajulia@vhebron.net