Correcting for single ascertainment by truncation for a quantitative trait.

Two methods for correcting for single ascertainment by truncation are compared. These methods are (1) conditioning on the phenotype of the proband and (2) conditioning on the event that the proband phenotype is greater than a threshold. The use of a constraint on model parameters is considered. The lack of robustness of this method to misspecifiction of the constraint has been demonstrated by Rao et al. It is noted that the constraint on model parameters used by Rao et al. is equivalent to an encoding of knowledge derived from a random sample, and an alternative representation of this information that has superior robustness properties is proposed.