SBC documentation

[melodiemonod]

Dear stan-dev team,

I have a question regarding the document on simulation based calibration:
https://mc-stan.org/docs/stan-users-guide/simulation-based-calibration.html

The document says:

To summarize, the following pseudocode computes the values for the test or for visualization in a histogram.

Inputs: M draws, J bins, N parameters, ranks r[n, m]
b[1:J] = 0
for (m in 1:M) {
  ++b[1 + floor(r[n, m] * J / (M + 1))]
}

Just before the ranks were indexed as r[n, k] where k was the number of parameters. Is there an error in the pseudo code that should read:

Inputs: N data sets, M draws, J bins, K parameters, ranks r[n, k]
For a given k:
b[1:J] = 0
for (n in 1:N) {
  ++b[1 + floor(r[n, k] * J / (M + 1))]
}

Many thanks for your help,

Best wishes,

Melodie

[melodiemonod]

Similarly I think:

For instance, if $J = 20$ and $M = 999$, then there are $1000$ possible ranks and an expected count in each bin of $(M+1) / J$

might actually be

For instance, if $J = 20$ and $M = 999$, then there are $1000$ possible ranks and an expected count in each bin of $(N+1) / J$

(this is visible from Figure 3)