Thank you very much for the replies, this worked well for me!
Another question regarding the estimate when having high pareto values: In the thread Interpret pareto k diagnostic - Modeling - The Stan Forums, you stated that “high Pareto k values indicate overoptimism”, which means that the (not trustable) LOO value suggests a better model fit than is actually the case. This means that it is enough to obtain the LOO manually for the better model because the LOO of the worse model can only get worse through the manual calculations, right?
Is there any reference I can use for this in my paper?