Situation in which the estimation of the ''interaction variance'' is the purpose of an experiment