On generalization of Bullen-Simpson's inequality (Q2496912)

The reviewer [Publ. Elektroteh. Fak., Univ. Beogr., Ser. Mat. Fiz. 602--633, 97--103 (1978; Zbl 0433.65014)] proved that if \(f\) is \(4\)-convex, then the dual Simpson's quadrature rule is more accurate than Simpson's quadrature rule; precisely: \[ 0\leq DS=\int_{0}^{1}f-\frac{1}{3}\bigl(2f(1/4)-f(1/2) + 2f(3/4)\bigr)\leq S=\frac{1}{6}\bigl(f(0)+4f(1/2) + f(1)\bigr)-\int_{0}^{1}f. \] The present paper extends this result to \(2r\)-convex functions by using Euler type identities, see \textit{Lj. Dedic, M. Matic} and \textit{J. Pecaric} [Panam. Math. J. 11, No. 2, 47--64 (2001; Zbl 1005.26016) and Bull. Belg. Math. Soc. -- Simon Stevin 8, No. 3, 479--504 (2001; Zbl 0994.26019)]. In particular if \(r\) is even, \(0\leq DS-T_{r-1}^{D}\leq S+ T_{r-1}^{S}\), where \[ T_{m}^{D}= -\frac{1}{3}\sum_{k=2}^{m}\frac{1}{(2k)!} \biggl(\frac{8}{2^{4k}}-\frac{6}{2^{2k}} +1\biggr)B_{2k}\bigl(f^{(2k-1)}(1)-f^{2k-1}(0)\bigr) \] and \[ T_{m}^{S}= \frac{1}{3}\sum_{k=2}^{m}\frac{1}{(2k)!}\bigl(1-2^{2-2k}\bigr)B_{2k}\bigl(f^{(2k-1)} (1)-f^{2k-1}(0)\bigr), \] the \(B_{2k}\) being Bernoulli numbers. There are many other results that are too technical to detail here.