Calculating WAIC for models with multiple likelihood functions

Questions
11

@OriolAbril
Also, I try the following -

idata.log_likelihood = xr.concat(
    [idata.log_likelihood[var_name] for var_name in var_names], 
    dim="aux_dim"
).sum("aux_dim")

Then I try to use the az.waic as follows =

az.waic(idata), which gives me the error -
AttributeError: 'DataArray' object has no attribute 'data_vars'

I then try -

idata.log_likelihood['samplenamehere'] = xr.concat(
    [idata.log_likelihood[var_name] for var_name in var_names], 
    dim="aux_dim"
).sum("aux_dim")

Followed by -
az.waic(idata, var_name='samplenamehere')

which gives the error -
KeyError: 'chain'

Is there some version mismatch issue?

I have the following package versions -
pymc3 = 3.9.3
theano = 1.0.5
arviz = 0.10.0
xarray = 0.16.1