-
-
Notifications
You must be signed in to change notification settings - Fork 407
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
HDI naming #2306
Comments
Thanks for reporting this. Probably a better name would be hdi_lb (lower bound) and hdi_up (upper bound). Do you want to submit a PR? |
A downside to the proposal of |
And what about something like |
In arviz, I think numbers would be shown to 3 decimal points, so if they're on unit scale, we have 4 digits plus decimal plus maybe a sign plus whitespace for 12-14 total characters for the column contents. If using a tuple, this is 14-16 characters. While In PosteriorStats.jl, we show 3 significant figures, so the tuple takes 12-14 characters, so I'm more strongly in favor of tuples there. Plus, table display in Julia truncates columns from the right instead of the middle like pandas, so I'm more strongly in favor of tuples in PosteriorStats. Here, I'm less certain. |
I'm in general not in favor of storing sequences in a cell of a data frame. However, I think this case is a good exception, so I vote in favor of the tuple. Having the column name be something like |
Hi,
I’m opening this because it’s not really clear why, in ArviZ’s output, the bounds for, say, a 94% HDI are named
hdi_3%
andhdi_97%
. Such names seem to imply that they correspond respectively to the 3rd and 97th percentiles of the posterior distribution, which would make them equal-tailed credible intervals if that were the case, but it isn’t necessarily. (HDIs and equal-tailed credible intervals don’t have to coincide.) Indeed, the following program shows a counterexample:Here is its output:
Is there maybe another reason for the naming? Or should it perhaps be reconsidered?
Thanks.
The text was updated successfully, but these errors were encountered: