Closed ahmad-shahi closed 1 year ago
Thanks for raising this. Both interval methods compute the level from both sides, but we're not handling this case correctly. Will fix soon
Hey @ahmad-shahi. Are you using level=[50]
with the conformal_distribution
method? We would appreciate if you can provide a code snippet.
Once model built, I use function predict model.predict(periods, level=[50]):
level
in this context means the width of the prediction interval, so in this case the interval uses the q25 and q75 as lower and upper limits.
Thanks for the clarification and good work. As advice, it is a bit confusing, level means confidence level, as users expect to return lower and upper bands as levels. You know better and it is just sharing my concern. Again, thanks for providing the good source for time series forecasting it is very user-friendly.
The reason is to be consistent with the forecast function from the forecast R package, where the lower and upper bounds of the interval are returned with similar namings (Lo 50
and Hi 50
).
Thanks for raising the issue, feel free to open another one if you encounter a problem. I'm closing this one.
The reason is to be consistent with the forecast function from the forecast R package, where the lower and upper bounds of the interval are returned with similar namings (
Lo 50
andHi 50
).Thanks for raising the issue, feel free to open another one if you encounter a problem. I'm closing this one.
Thanks for the good work and keep on.
When I use level 50, why it returns low and high for 50. 50 should be median and should return median. is it?