To investigate the impact of different GRM thresholds for sample elimination on the GCTA estimates, we varied the GRM threshold from .01 to 1.0 (i.e., all individuals included) for each of the five clinical phenotypes. Figure 1 provides illustrative results from these analyses for the Alcohol Consumption factor; plots for the remaining four clinical phenotypes are given in the Supplementary material (Supplemental Figures S21 to S24). As can be seen in the plot, the GCTA estimates cluster between 10% and 20% when the GRM threshold is less than .10. When the threshold is increased to .50, in which case first-degree relatives are being included, the estimate increases to 30%, and it further increases to 60% when the threshold is set to its boundary value of 1.0, in which case MZ twin pairs are also included. At the higher GRM thresholds, the GCTA estimate is being driven by the relationship of phenotypic similarity to genetic relatedness among closely-related individuals (i.e., it is essentially biometric). At the lower GRM thresholds, GCTA is estimating the proportion of phenotypic variance accountable by all of
