As I understand it,
Panel Data is already represented as a GroupedDataFrame
where you groupby
the unit/location/etc, and you have a time column.
We should think about adding lags overloads for that in packages that do lags.
For years I was confused by panel data, until I realized that.
I think encouraging thinking of it as 1 particular case of groupby
/GroupedDataFrame
is good.
Econometricians have played us for fools.