The parameters of the Normal-Gamma prior are not very reproducible, especially when a LogNormal sampling distribution is used. Some possible solutions:
Optimise Gamma parameters based on coefficient of variation, then select Normal parameters with prior predictive distribution
Plot marginals
Plot marginals of LogNormal mean and coefficient of variation
Directly optimise Normal parameters by computing mean in real space quantiles with delta method
The parameters of the Normal-Gamma prior are not very reproducible, especially when a LogNormal sampling distribution is used. Some possible solutions: