On uniform confidence bands for conditional effects, a simple approach for small-N inference, refining the "no anticipation" assumption, and the LP-DiD estimator for staggered designs.
I like the LP-DID framework since such approach would be way more simpler and cleaner than other approaches, but I wonder whether this means the outcome variable of the LP-DiD should be detrended (if there is a unit root) or deseasonalized as well, similar to the original implementation and issues discussed in the LP literature
I like the LP-DID framework since such approach would be way more simpler and cleaner than other approaches, but I wonder whether this means the outcome variable of the LP-DiD should be detrended (if there is a unit root) or deseasonalized as well, similar to the original implementation and issues discussed in the LP literature