Alternatively, you can just crank up the amount of data you are modeling (i.e., the number of observations). This will make the various approximations “less approximate”:
observed_data = np.random.normal(loc=true_mean_of_generating_process,
scale=true_sigma_of_generating_process,
size=100_000)
...
az.plot_ppc(idata, group="posterior", num_pp_samples=100)
