##
**The significant-digit phenomenon.**
*(English)*
Zbl 0833.60003

The paper is a survey article on scale-invariance and base-invariance. In order to explain the significant-digit phenomenon (Benford’s law) the author defines a natural domain \(\mathcal A\) on the positive reals which is generated by infinite basis sets \(S_{a,b} = \bigcup^\infty_{i = - \infty} [a 10^i, b 10^i)\), \(1 \leq a < b \leq 10\). On \(\mathcal A\) significant-digit laws to base 10 can be studied. A probability measure \(P\) is called scale-invariant if (i) \(P(S) = P(\alpha S)\) for all real \(\alpha > 0\) and all \(S \in {\mathcal A}\). A probability measure \(P\) is called base-invariant if (ii) \(P(S_{1,b}) = P(S^{1/k}_{1,b})\) for all positive integers \(k\) and all basis sets \(S_{1,b}\), where \(S^{1/k} = \{x^{1/k} : x \in S\}\) [cf. the author, Proc. Am. Math. Soc. 123, No. 3, 887-895 (1995; Zbl 0813.60002)]. Both scale-invariance and base-invariance imply Benford’s law, essentially. The special role of the constant 1 is discussed.

Reviewer’s remarks: 1. Under base-invariance of a probability measure \(p\) the defining relation (ii) holds for all \(S \in {\mathcal A}\), not only for basis sets. 2. For the definition of scale-invariance of a probability measure \(P\) it suffices that the defining relation (i) is satisfied only for basis sets \(S_{1,b}\). Then it holds for all \(S \in {\mathcal A}\). After this, the definitions and properties of scale-invariance and base- invariance would look more alike. 3. The construction of \(\mathcal A\) has the advantage of avoiding random variables in the definition of scale- invariance and base-invariance but the disadvantage of not explaining the simultaneous validity of Benford’s law to several bases.

Reviewer’s remarks: 1. Under base-invariance of a probability measure \(p\) the defining relation (ii) holds for all \(S \in {\mathcal A}\), not only for basis sets. 2. For the definition of scale-invariance of a probability measure \(P\) it suffices that the defining relation (i) is satisfied only for basis sets \(S_{1,b}\). Then it holds for all \(S \in {\mathcal A}\). After this, the definitions and properties of scale-invariance and base- invariance would look more alike. 3. The construction of \(\mathcal A\) has the advantage of avoiding random variables in the definition of scale- invariance and base-invariance but the disadvantage of not explaining the simultaneous validity of Benford’s law to several bases.

Reviewer: P.Schatte (Freiberg)

### MSC:

60A10 | Probabilistic measure theory |

60E99 | Distribution theory |

11K16 | Normal numbers, radix expansions, Pisot numbers, Salem numbers, good lattice points, etc. |