Let X be a Banach space. Consider the family I(α) of linear continuous operators from X into itself, depending on the parameter α ≥ 0. Suppose that I(α1+α2)=I(α1)I(α2)∀α1, α2 ≥ 0 and I(0)=I (semigroupal property). Such a semigroup is said to be strongly continuous for α > 0, if limh→0I(α+h)x =I(α)x, ∀x ∊ X.

P. Erdös and G. Szekeres [1] proved that from any points in the plane one can always choose n + 1 of them which are the vertices of a convex polygon, thus answering a question due to Miss Esther Klein (who later became Mrs. G. Szekeres).

This paper was inspired by [4]. The main result there suggests a close relationship between injective hulls of C* algebras as studied in [2] and [3] and rational completions as studied in [1]. We shall prove an analogue of Theorem 1 in [3] for rational completions. The latter theorem states that the injective hull of the algebra C(X) of all complex valued functions on the compact Hausdorff space X is the algebra B(X) of all bounded Borel functions modulo sets of first category.

We consider the following functional equation

1

where ƒ = ƒ(x, y) is a real-valued function of two real variables x, y on the whole xy-plane and t is a real variable.

With regard to the geometric meaning of (1), the equation is called the “square” functional equation.

In this paper we use some known transformations available in the Theory of Multiple Integrals to give straightforward, simpler, and elegant proofs of some distributions of ordered values from Burr and beta distributions. The exact distribution (under the null hypothesis) of Wilks' ∧ criterion is obtained by considering it as a certain minimum value distribution problem.

The transforms discussed here are the quasi-Hausdorff (Theorem 1), the [J, ƒ(x)] (Theorem 2) and the Borel integral means (Theorem 3). We are concerned here with whether or not the limit-points of these transforms of bounded sequences form connected sets. Such a set is one which cannot be decomposed into the union of two disjoint nonempty open sets.

It is shown that if a chainable continuum has exactly two arc components, then one of them is an arc and the other is a half-ray.

A topological space X is called maximally resolvable if it admits a largest possible family of pairwise disjoint, “maximally dense” subsets. More precisely, if Δ(X) denotes the least among the cardinal numbers of the nonvoid open subsets of X, then X is maximally resolvable if it has isolated points or there exists a family {Rα}α < Δ(X) of subsets of X, called a maximal resolution for X, such that ∪{Rα | α < Δ(X)} = X, Rγ ∩ Rδ==ϕ if γ ≠ δ, and, for each a and each nonvoid open subset V of X, the cardinality of Rα ∩ V is not less than Δ(X).

Let Σ be a left amenable semigroup, and let {Tσ: σ ∊ Σ} be a representation of Σ as a semigroup of positive linear contraction operators on L1(X, , p). This paper is devoted to the study of existence of a finite equivalent invariant measure for such semigroups of operators.

In this brief exposition we collect several results on rings of formal power series with coefficients from a field or a ring with some special properties. The results that are catalogued below are mostly algebraic in nature.

Let (G, ρ) be a finite connected (undirected) graph without loops and multiple edges. So x, y being two elements of G (vertices of the graph (G, ρ)), 〈x, y〉 ∊ ρ means that x and y are connected by an edge. Two vertices x, y ∊ G have the distance μ(x, y) equal to n, if n is the smallest number with the following property: there exists a sequence x0, x1, …, xn of vertices such that x0 = x, xn = y and 〈xi-1, Xi〉 ∊ ρ for i = 1, …, n. If x ∊ G, we put μ(x, x) = 0.

The purpose of this note is to provide integral inequalities which are related to Hardy's ([2] and [3, Theorem 330]). This latter result we state as

Theorem 1. Let p>1, r≠1, and ƒ(x) be nonnegative and Lebesgue integrable on [0, a] or [a, ∞] for every a>0, according as r> 1 or r< 1. If F(x) is defined by

1

then

2

unless f≡0. The constant is the best possible.

In [1] we constructed a (G, σ)-space and determined a condition which is both necessary and sufficient for the (G, σ)-product of the vectors v1, v2, …, vn to be zero. The purpose of the present paper is to give a criterion for v1 Δ v2 Δ … Δ vn to be zero, in the particular case when V is a unitary space and the group G belongs to a special class of groups which we shall define below.

We follow [6] and [7] for all terminologies. The purpose of this note is to prove

Theorem 1. Let X and Y be any two integer-compact spaces. The following are equivalent:

1. (1) X is homeomorphic to Y.

2. (2) C(X, Z) and C(Y, Z) are isomorphic as rings.

3. (3) C(X, Z) and C(F, Z) are isomorphic as lattices.

4. (4) C(X, Z) and C(Y, Z) are isomorphic as p.o. groups.

5. (5) C(X, Z) and C(Y, Z) are isomorphic as multiplicative semigroups.

It is known (cf. M. A. McKiernan [6]) that the only measurably bounded solutions ƒ of the equations

1

where x ∊ Rn, t ∊ R, αi(i= 1, …, m) span the space , and Σi ∊ 1 μi ≠0 for any I ⊂ {1, …, m}, are polynomials. The degree of these polynomials and the dimension of the solution space can be estimated by numbers depending on m and n. (For estimates and other details concerning equations (1) see see [1], [2], [3], [4], [5], [6].)

Paul Halmos expressed [3, p. 110] the general dissatisfaction with the usual proofs of this famous and important theorem. They all make it seem like an accidental product of a computation. A more conceptual proof was devised by N. P. Dekker. In spite of the elegance of his proof, the one offered below may have some claim to be regarded as "the reason the theorem is true".

We shall be concerned with the following problem. Let a11, …, ann be complex numbers and λ1, …, λn nonnegative real numbers. Under what conditions does there exist an n × n complex matrix A with a11, …, ann as principal elements and λ1, …, λn as singular values? This problem has been suggested in [3] but, to our knowledge, has not yet been solved.

L'emploi de comme estimateur de la moyenne est très fréquent en statistiques pour plusieurs bonnes raisons. Ainsi pour une densité régulière de la classe exponentielle, il est l'estimateur sans biais pour la moyenne, à variance minimale. Mais il perd cette importante propriété pour la classe des densités tronquées étudiées ici: cet article démontre que lorsque la taille de l'échantillon est suffisamment grande, il faut préférer Δ = [(mini=1, …, nXi + maxi=1, …, nXi/2] à .

Let Q be a finite partially ordered (by ≤) set with universal bounds O, I. The height function h of Q is defined by the rule: h(x) is the maximum length of a chain from O to x. Let h(I)=n. Suppose that for each k≥0, there exist positive integers a(k) and b(k) such that all elements of height k

1. (i) are covered by a(k) elements of height k+1;

2. (ii) cover b(k) elements of height k—1.

Then we call Q a U-poset. Call a subset S of a partially ordered set an antichain if no two elements of S are comparable.

In this note we give a very elementary proof of a result of Meyer's [1]. Let r(A) be the rank of the matrix A.