# When is it acceptable to compute (conditional) subset-averaged coefficients?

I'm running an ecological study and I have 4 dependent variables (DVs) that I would like to explain (my interest thus lies in inference and not in prediction). For each one of these variables, I built a set of a priori candidate models based on ecologically relevant hypotheses. Following Burnham & Anderson (2002), I selected the most parsimonious models on the basis of AICc (with delta AICc < 4) and would like to average their coefficients using the MuMIn package in R.

Now, I understand the difference between (unconditional) full model averaging and (conditional) subset-based model averaging but I'm not sure I understand when to use one or the other. Is computing unconditional coefficient estimates just a way of being more conservative? If I have reasons to believe that the relationships between some of my IVs and the DV may be weak, should I compute conditional averages? Should I report both in a scientific article?
This question already adresses this question but the answers focus on the difference between these types of multimodel inference and not on when it is appropriate to choose one over the other. Thanks for your help!

• @fanoue I'm sorry no one has answered this. I posted a related question where I also express an opinion. Jan 12 at 20:34