The most striking feature revealed by our linkage analysis of the quantitative trait Q1 is that, with one possible exception, none of the five signals seen in the Luhya sample colocalize with any of the four signals found in the Yoruba sample. The exception is the signal seen on chromosome 8 at CDCA2, although only the Yoruba signal is highly significant. Had we analyzed each of the three Luhya pedigrees and each of the two Yoruba pedigrees separately, we would have found family-specific signals, all of which would have been stronger than those reported in Table 1 because of our pooling of pedigrees from the two ethnic groups. Clearly, for private polymorphisms or in cases of locus heterogeneity (see, e.g., Morton [7] for the first published example of locus heterogeneity documented by a linkage analysis), linkage analysis of individual families often is preferable to combining all families from a single ethnic group, as we originally did. Indeed, Culverhouse et al. [8], analyzing the unrelated individuals data set, also found that pooling populations could obscure a strong signal even when Population was included as a covariate.
