I've reconstructed states for a group of organisms using the HiSSE and BiSSE models. The AICc values are very close, but the two models generate contrasting ASRs. So I think I need to combine them using model-averaging (with Akaike weights). But I'm not sure how to go about it and I can't find an example in the literature.

Can I model-average likelihoods for each state at a given node? Or do I model-average the parameters for BiSSE and HiSSE and then use those parameters to generate the ASRs?

And, if I model-average the parameters for the two models, how should I deal with HiSSE having more parameters than BiSSE (i.e., for hidden states)?