[R] Tukey's one degree of freedom for nonadditivity?
spencer.graves at pdf.com
Thu Apr 3 22:11:54 CEST 2003
Is there code available to decompose interactions involving at least
one nominal factor with more than 2 levels as described, e.g., by Tukey
or by Mandel (1971, Technometrics, 13: 1-18)?
E(y[i,j]) = mu0 + a[i] + b[j] + c*a[i]*b[j],
estimating a, b, and c so sum(a) = sum(b)= 0. Mandel essentially
describes a singular value decomposition of the interaction.
More information about the R-help