We have noticed some issues with KP F-stats for a few cases where the first-stage is exactly identified. We will write a more detailed post on this later (this also applies to ivreg2 and ivreghdfe in stata).
There is a knife-edge case for the formula for the F-stat in ranktest which needs to be considered.
This is cross-posted from https://github.com/FixedEffects/FixedEffectModels.jl/issues/150
We have noticed some issues with KP F-stats for a few cases where the first-stage is exactly identified. We will write a more detailed post on this later (this also applies to
ivreg2andivreghdfein stata).There is a knife-edge case for the formula for the F-stat in
ranktestwhich needs to be considered.