I have a 2x3 factorial design, and wish to explore specific hypotheses with planned (orthogonal) contrasts. Some people recommend that in order to specify the contrasts, this design can be replaced with a 6-level single-factor design. However, the advantages (besides the simplicity of constructing the subsequent contrasts, or perhaps estimating p values etc. if done in a linear mixed model) versus disadvantages/problems with this approaches are unclear to me.
Are the two designs equivalent when contrasts are orthogonal? Can this be done regardless of whether the interaction term is significant or not? Is there a good reference for this approach?