I ran some sdid regressions and it seems that the "projected" method is much slower than the "optimized" method, while the sdid accompanying paper suggests that the "projected" method can be significantly faster.
Do you have any idea why this is the case and if there are ways to speed up the computation?
Here is the code I run (the data file is attached):
gen Z08_log = log(Z08)
timer on 1
sdid log_RS_filo cod_insee year treated, covariates(H12 E12 Z08_log, projected) vce(bootstrap) reps(10) seed(123)
timer off 1
timer on 2
sdid log_RS_filo cod_insee year treated, covariates(H12 E12 Z08_log) vce(bootstrap) reps(10) seed(123)
timer off 2
timer list
Hi,
I ran some sdid regressions and it seems that the "projected" method is much slower than the "optimized" method, while the sdid accompanying paper suggests that the "projected" method can be significantly faster.
Do you have any idea why this is the case and if there are ways to speed up the computation?
Here is the code I run (the data file is attached):
Thank you very much!
test.csv