On an inequality arising from population genetics (Q1072556): Difference between revisions

Let \(n\) be some positive integer, let \(p_1,\dots,p_n\in [0,1]\) such that \(\sum_i p_i=1\) and let \((f_{ij})\) be some \(n\times n\)-matrix over \([0,\infty)\). Further assume \((p'_1,\dots,p'_n)\neq (p_1,\dots,p_n)\) where \(p'_i:=p_i\sum_j f_{ij}p_j/\Phi\) for all \(i=1,\dots,n\) and \(\Phi:=\sum_{i,j} f_{ij}p_ip_j>0\) and put \(\phi (t):=\sum_{i,j}f_{ij}(p_i+t(p'_i-p_i))(p_j+t(p'_ j-p\_j))\) for all \(t\in \mathbb R\). The inequality \(\phi(0)<\phi(1)\) is called ``fundamental theorem of natural selection''. In the paper a short and elementary proof of the result \(\phi(0)<\phi(t)\) for all \(t\in (0,1]\) originally obtained by \textit{J. Rosner} and \textit{W. Timischl} in [Sitzungsber., Abt. II, Österr. Akad. Wiss., Math-Naturwiss. Kl. 188, 247--254 (1979; Zbl 0452.92016)], is given. Moreover, generalizing a previous result obtained by \textit{W. Timischl} and the author in [Z. Angew. Math. Phys. 31, 538--541 (1980; Zbl 0463.92012)], sufficient conditions are provided which guarantee strict monotonicity of \(\phi\) along \([0,1]\).