Most of our discussion has addressed computational issues. However, there is one methodological issue that deserves a brief comment. The online documentation of PLINK 1.07 weighed the pros and cons of its permutation procedure for within-family analysis of quantitative traits (QFAM) with respect to the standard quantitative transmission disequilibrium test (QTDT) [35]. It pointed out that likelihood-based QTDT enjoyed the advantages of computational speed and increased statistical power. However, a comparison of statistical power is only meaningful if both procedures are anchored to the same Type 1 error rate with respect to the null hypothesis of no linkage with a causal variant, and Ewens et al. has shown that the QTDT is not robust against certain forms of confounding (population stratification) [36]. On the other hand, the validity of a permutation procedure such as QFAM only depends on the applicability of Mendel’s laws. When this nicety is combined with the vast speedup of permutation in PLINK 1.9, a given user may now decide to rate QFAM more highly relative to QTDT when considering available options for within-family analysis.
