Re: [Gretl-users] Fixed Effect crashed when I tried to include over 5000 entity dummies
Am 10.02.2019 um 13:04 schrieb Fred Engst:
2. I didn’t realize that is how gretl does the fixed effect, for it
reported LSDV R-squared, which I thought means it included the contribution of those
dummies.
Please see the heading "R-squared in the fixed effects model" in the
user's guide ch. 20 (p.171 in my A4-format copy from Dec 2018).
3. I really don’t know much or thought much about “fixed effect”
tobit. I just thought by adding those dummies I don’t have to demean those dependent
variables. Otherwise they will lose their non-nagetive nature, making tobit a misfit. Now
I realized that to estimate over 5000 parameters is next to impossible anyway.
It seems that if I reduce the detentions of those “fixed effects” to a few
characteristics, say less than 10, then the tobit model should work.
Not sure what you mean by reducing the fixed effects to 10 or by
detentions in this context, but again, an unbiased parametric fixed
effects Tobit estimator does not exist. (But it could be that the bias
is not always severe, don't know about that.)
good luck,
sven