Handle multiple observations from same id and wave gracefully

[jacob-long]

This is very cool. Working on a package for cleaning panel data; I'll add a function to convert the result to a panel_data object.

Quick note: if I give it multiple observations per id, it lets me convert to panel_data without error, but then lag doesn't work properly pic.twitter.com/K7eVyHeXNr

— Nick HK (@NickCHK) May 21, 2019

Interesting use case. I should probably think about a way for users to communicate additional grouping levels like this. I wonder if the "right" way is to have an ID that is a unique combination of student and subject.

— Jacob Long (@jacobandrewlong) May 21, 2019

That doesn't quite work because if I want, say, last period cumulative GPA but you didn't take that subject last term it will give a NA incorrectly. Lag is poorly defined here, should probably give an error on declaring the object

— Nick HK (@NickCHK) May 21, 2019

That might be the safest method to avoid silently giving wrong results. Then the user can calculate group mean statistics at the student level before creating the panel_data object using id-subject as the panel ID, forcing the user to make these decisions explicitly

— Jacob Long (@jacobandrewlong) May 21, 2019