[Gretl-users] Re: A query re the equation version of quantile regression.
On Mon, 11 Jan 2021, Sven Schreiber wrote:
Am 11.01.2021 um 17:37 schrieb Brian Revell:
> I estimated 0.1 0.2 0.5 and 0.9 quantiles. THe equation constant and
> slope appeared to fall between the 0.1 and 0.2 quantiles. I first
> compared the equation coefficients with the 0.5 quantile estimates-but
> there was no similarity with the coefficients
> Hope that is clear.
OK, when I estimate several quantiles at once (didn't remember that that
was possible!) I can confirm that when I choose (from the model window)
File -> View as equation there is a _single_ equation shown which has
mysterious coefficients, meaning they do not or only partly match
coefficients in the full output.
Could be a bug alright, or else I agree it needs to be explained! Thanks
for reporting it.
"View as equation" should not be enabled for multi-tau quantreg
estimates. That's now fixed in git. What was shown was actually an
initial set of OLS estimates, which are just for internal use.
Allin