Relevant dyads are pairs of contiguous states or pairs of states including
at least one major power. They are argued to be the population of dyads at
risk of international conflict and are increasingly commonly used as the ca
ses analyzed by conflict researchers. Does reliance on relevant dyads (a no
nrandom sample of all dyads) introduce threats to valid inference? The auth
ors argue that relevant dyad usage might introduce related problems of meas
urement error and selection bias and investigate whether there is evidence
of such potential problems existing in actual relevant dyad data sets. Resu
lts show evidence for both types of potential problems, but neither problem
substantively affects estimates of conflict relationships. Therefore, rele
vant dyad usage is not especially objectionable due to either of these prob
lems, and retrospective sampling might be an even more profitable course fo
r future research.