Currently I'm regressing two values (call it x and y) that are time series ordered pairs. I was wondering if there was a way to add a representation of this third dimension (time) to the scatter plot regression. One thought was to have the individual x,y points fade as they get older.
Not necessary to get caught up in the details, but for context:
I am regressing values of financial instruments. As time changes, the relationship tends to adjust to "new regimes." It would would be nice to see if the current pair value is simply off from the model due to a short term market mispricing, or if the current deviation from the model mean is due to the relationship shifting in to a new regime (ie. many of the recent points are clustered in the same area away from the model mean. This implies that going forward the model will adjust towards the more recent points, as opposed to the pair reverting back towards the model mean.)
Any and all suggestions welcome (even if conceptual instead of technical.) Thanks all.
Bookmarks