Prediction interval vs. confidence interval

[MarleneBs]

I discussed with my supervisor that a confidence interval and a prediction interval are not the same thing.
However, in the documentation for the plot function of the forecast object, the whole thing is used synonymously. In the description it says confidence interval and the parameters to be set are called forecast interval. Which of the two is now displayed in the plot function for the forecast object?

Translated with www.DeepL.com/Translator (free version)

[codingWhale13]

That's a good point. I'm not sure myself but maybe this helps a bit:
As you can see here, plot just uses the quantile method of Forecast. For example, for a 90% interval, the 0.05 and 0.95 quantiles are used for plotting.

[MarleneBs]

Thank you very much

[MarleneBs]

I have to get back to you,
I don't really understand it. In all literature there is a separation between prediction interval and confidence interval. Why is it used as a synonym in the documentation:
https://ts.gluon.ai/stable/_modules/gluonts/model/forecast.html#Forecast.plot

[MarleneBs]

I think i have found a definition that is equivalent to your statement
thank you

[codingWhale13]

You're right, they are two different things. As @jgasthaus was saying, in the GluonTS context, we are concerned with the prediction interval.

I updated the documentation in #2373 (Might take a while to update on the website as well).

[MarleneBs]

Thank you very very much, this helps me a lot