2 way fixed effect Panel data with Julia?

[jlperla]

@schrimpf A thought if we do cover any panel data.... people might be sold if the speed difference is huge, and the interface is easy.

• https://github.com/matthieugomez/FixedEffectModels.jl
• https://github.com/jmboehm/RegressionTables.jl for the graphs...

My basic thought is that:

• Data wrangling could be done with https://s3.amazonaws.com/assets.datacamp.com/blog_assets/Python_Pandas_Cheat_Sheet_2.pdf (or tell people about dplyr and tidy)
• Then in julia can always load the pandas trivially with https://github.com/JuliaPy/Pandas.jl
• Then use the FixedEffectModels.jl for blinding speed...

[jlperla]

Also, https://github.com/matthieugomez/InteractiveFixedEffectModels.jl/blob/master/README.md for interactive fixed effect. Applied macro guys use this... not sure about others.

A nice "big" project for a student would be to try to implement the FixedEffectModels.jl test with a matrix-free Krylov method (e.g. https://github.com/JuliaSmoothOptimizers/Krylov.jl is one, It seems to me that a REALLY big 2-sided fixed effect could make use of the fact that there are only 2 nonzeros in each row, which makes the matrix-vector inner product very efficient?