Differences between true and observed parallax: effect of removing observed negative parallaxes and those above a given relative error. We start with a representative subsample of 1 million stars (top figure) and truncate it according to the apparent relative parallax precision. Top: complete sample. Mean difference ϖ − ϖ_True is 1.55 × 10⁻⁵ mas. Middle: retaining only objects with positive parallaxes and . The mean difference ϖ − ϖ_True is 0.164 mas. Bottom: retaining only objects with positive parallaxes and . The mean difference ϖ − ϖ_True is 0.026 mas.