QTDT - Permutations

In small samples, in selected samples or when the variance model is incomplete, estimates of the variance components may be biased. QTDT can calculate empirical p-values for the within family component of association using permutations. Use the -m option to specify the number of Monte-Carlo permutations to perform.

For example to run 1000 permutations for the orthogonal model (-ao) and modelling environmental, polygenic and additive major locus effects (-wega) run:

 prompt> qtdt -ao -wega -m1000

Another attractive feature of using Monte-Carlo permutation to estimate p-values, is that they can provide a global p-value with a built-in adjustment for multiple testing. To do this, QTDT permutes transmission scores for all markers simultaneously and notes the most significant association observed in each replicate. The most significant association in the actual data is then compared to this reference distribution. This strategy is based on that of McIntyre et al (2001).

Caution: On smaller computers evaluation of permutations can be slow!