Well, maybe the way to handle it is to use the normal model for december but then multiply the resulting vector by an inverting matrix....(n/g), luckily, my correlation equations are not tensors (g).