Recently there is a lot of discussion about global forecasting models and their advantages/disatvantages compared to local models. My question is not about this comparison. I want to know the difference, what it means in mathematical and statistical terms. I could not find any helpful post or scientific articles on this topic.
In terms of data preparation, you simply change from a wide (column-wise) format to a long (panel) format. Does this imply a forecast of time series groups?
In mathematical terms, you are fitting one function to all time series. Is there anyone, who can give an example for this?
In terms of statistics I tend to get really confused about this topic. There are a lot assumptions and heuristic methods, to get a better performance in global models. I heard about clustering, but all my efforts to create useful clustering or filtering.
For any hint in towards the right literature or some real world examples, I would be thankful!