For studying certain problems in mathematical population genetics one may introduce nonassociative linear algebras specifying the probability distribution of offspring genotypes given the genotypes of the parents. In this paper it is pointed out that it may be useful to regard these algebras as vectors in a larger space. The set of baric algebras with a fixed weight function and the set of genetic algebras with a fixed canonical basis are shown to form flats in this space. Possible choices of a suitable basis are discussed for the subspace of algebras corresponding to linked autosomal loci. It is shown that there exists in some situations a particular basis of algebras with genetic interpretation such that all other algebras of possible genetic significance can be expressed as convex combinations of the basis elements.