Repairing the proof of a classical differentiation result (Q1332612)

Let \(E\) be a Lebesgue measurable subset of \(\mathbb{R}\) and let \(f: \mathbb{R}\to \mathbb{R}\) be a measurable function. The paper contains some new proofs of the following \textit{J. M. Ash's} [Trans. Am. Math. Soc. 126, 181-199 (1967; Zbl 0164.359)] results: 1. Suppose \(|\sum^ n_{i= 0} A_ i f(x+ a_ i t)|= O(1)\) for all \(x\in E\). Then \(f\) is bounded in a neighbourhood of almost every point of \(x\in E\). 2. Suppose \(\sum^ n_{i= 0} f(x+ a_ i t)= O(t^ \alpha)\) for all \(x\in E\). Let \(a\in \mathbb{R}\). Then \[ \sum^ n_{i= 0} A_ i f(x+ (a_ i- a)t)= O(t^ \alpha) \] for almost every \(x\in E\). If ``\(O\)'' is replaced by ``\(o\)'' in the assumption and in the conclusion then the resulting assertion is also true. Earlier proofs of these and related results made use of the measurability of the sets like \[ E_ j= \left\{x\in E: \Bigl| {f(x+ h)- 2f(x)+ f(x- h)\over h^ 2}\Bigr|< j\text{ for all } 0< | h|< {1\over j}\right\}. \] There is an example of a measurable function for which the set \(E_ 2\) is not measurable.