Special treatment of observed data for prior predictive checks

[aloctavodia]

I have observed that people often uses plot_ppc with group prior and compares the prior predictive to the observed data. I think this is a potentially risky practice. To reduce the potential risk we should automatically treat observations differently when group is prior instead of posterior. I have 3 ideas.

• don't show observed data by default.
• show them with a warning message.
• show bootstraped onservatios by default.

[ahartikainen]

I think not showing them by default is a good idea.