One nice feature of R's step() function is that it can increase the search space to include interactions/higher order terms for the effects that were already included. Question: does this fit in the affine and/or quadratic frameworks? It seems like it should, since the inclusion of the linear/main effects is already being tracked and conditioned on.
One nice feature of R's step() function is that it can increase the search space to include interactions/higher order terms for the effects that were already included. Question: does this fit in the affine and/or quadratic frameworks? It seems like it should, since the inclusion of the linear/main effects is already being tracked and conditioned on.