What diagnostics to learn if the model has too little data?

I am currently training models on small datasets with up to 200 points. But in some cases almost all the points we be repeated so that there are just one or two unique points. If there a diagnostic or some output I can check the quality of the posterior prediction and of specific parameters corresponding to those provided by OLS regression ? Also if there is any pointers regarding the robustness of the model to say an outlier.

I am currently use arv.viz summary(posteriror) to evalute the model - but I am not sure how to interpret all its information.

Many thanks !

I often look at the posterior variance of the parameters I’m estimating. In models that learn an observation noise level, you can compute something like a signal-to-noise ratio by comparing the posterior variance to the learned observation noise.