In this paper is given a Bayesian solution of the Fieller-Creasy problem in the case, when the variance of populations are not assumed equal. More precisely, let $X\sb 1,...,X\sb n$ be i.i.d. rv’s distributed according to $N(μ\sb x,σ\sb x)$ and $Y\sb 1,...,Y\sb m$ be i.i.d. rv’s distributed according to $N(μ\sb y,σ\sb y)$, $σ\sb x\ne σ\sb y$, $ξ=μ\sb x/μ\sb y$. The author gives the exact a posteriori distribution $π$ ($ξ\vert X,Y)$. The results are illustrated on practical and artificial data sets.
J.Antoch