For a 9th magnitude star like TYC 1718 2095, the parallax is given as
4 +/- 14 milliarcsceonds. When the error bars are this large, is
estimating the distance even possible? I guess I am asking whether a
range of 28 milliarcseconds really does capture the limits on the
distance?

And as an aside, does anyone happen to know the source of the error?
Since it blows up for fainter stars, I would guess that a principle
component must be Poisson noise. But I don't see why that would
affect centroiding so dramatically.

Ben Davies