Epistasis, is the interaction between alleles from two or more loci determining complex traits, and thus plays an important role in the development of quantitative traits of crops. In mapping studies of inbreeding species epistasis is usually defined as the interactions between quantitative trait loci with significant additive gene effects. Indeed, in many studies, genes with small effects do not come into the final model and thus the total epistasis interaction effect is biased. Many loci may not have a significant direct effect on the trait under consideration, but they may still affect trait expression by interacting with other loci. In this paper the benefits of using all loci, not only the loci with significant main effects, for estimation of the epistatic effects are presented. The particular examples are with doubled haploids lines and so are restricted to homozygotes and thus additive genetic effects and additive 9 additive interactions. Numerical analyses were carried out on three populations of doubled haploid lines of barley (Hordeum vulgare L.): 120 doubled haploid lines from the Clipper 9 Sahara 3771 cross, 145 doubled haploid lines from the Harrington 9 TR306 cross and 150 doubled haploid lines from the Steptoe 9 Morex cross. In total, 157 sets of observations were analyzed and altogether 728 pairs of loci were observed for the three datasets.