Hi, I'm trying to use rlm function in an unequally replicated 2x3 factorial experiment all combinations of 3 temperatures by 2 amounts.(cDNA microarrays) For each gene I have 6 log ratios, and can do a standard factorial experiment analysis, giving a grand mean, a main effect of amount, and 2 degrees of freedom for the main effect of temperature, a linear effect and a quadratic effect, These are the main effects, and the two degrees of freedom interaction effects, see matrix below. Here is a little matrix of coefficients for all these terms. mean + + + + + + amount - + - + - + linear temp - - 0 0 + + amount temp + + -2 -2 + + amount.linear + - 0 0 - + amount.quad - + 2 -2 - + This becomes and orthogonal transformation of the 6 observations by normalizing the rows appropriately. Could you please suggest me how to include this matrix design into rlm function? Thanks in advance. SV