Hello everyone,
I run a threshold regression following the Hansen(1999) paper and the code from Wang(2016) as follows:
The code I use is
However, my panel data contains only 15 countries (15 clusters) which is a low number. To correct for that I am interested in using something like Wild bootstrapping to produce bias-adjusted clustered standard errors.
To do so, I tried boottest and clustse and both of them seem to be incompatible with XTHREG.
Any suggestions what to do to deal with the fact that I have very few clusters?
I run a threshold regression following the Hansen(1999) paper and the code from Wang(2016) as follows:
The code I use is
xi: xthreg DD F1 G1 H1 I1 J1 i.year, rx(E1) qx(F1) thnum(1) grid(400) trim(0.1) bs(300) cluster(country)
To do so, I tried boottest and clustse and both of them seem to be incompatible with XTHREG.
Any suggestions what to do to deal with the fact that I have very few clusters?
Comment