The inverse of h2_observed(): h2_lat = h2_obs * p (1 - p) / z^2
(Dempster & Lerner 1950). This is the direction a user needs when a linear model
was fit to a 0/1 trait and the estimate has to be reported on the scale the
threshold literature uses.
Value
heritability on the liability scale. A warning fires when the pair implies a liability h2 above 1: the two numbers are then not consistent with the threshold model.