Approximation theorems for strongly mixing random variables (Q788385)

Let (S,D) be a Borel space and (\(\Omega\),F,P) be a probability space. The main purpose of the paper is to derive some sharp ''approximation theorems'' (Theorems 2,3 and 4) which allow us to approximate S-valued dependent random variables satisfying the ''strong mixing condition'' by independent ones. The closeness of this approximation is measured by the following measures of dependence: \[ \alpha({\mathfrak A},{\mathfrak B})=\sup | P(A\cap B)-P(A)P(B)|,\quad A\in {\mathfrak A},\quad B\in {\mathfrak B} \] \[ \rho({\mathfrak A},{\mathfrak B})=\sup | Corr(f,g)|,\quad f\in L^ 2({\mathfrak A}),\quad g\in L^ 2({\mathfrak B}) \] \[ \beta({\mathfrak A},{\mathfrak B})=\sup 1/2\quad \sum^{I}_{i=1}\sum^{J}_{j=1}| P(A_ i\cap B_ j)-P(A_ i)P(B_ j)| \] where this last sup is taken over all pairs of partitions \(\{A_ 1,...,A_ I\}\) and \(\{B_ 1,...,B_ J\}\) of \(\Omega\) such that each \(A_ i\in {\mathfrak A}\) and each \(B_ j\in {\mathfrak B}\); \({\mathfrak A}\) and \({\mathfrak B}\) are any two \(\sigma\)-fields, \({\mathfrak A}\subset F\), \({\mathfrak B}\subset F\). The proofs of the mentioned theorems are essentially based on the approximation theorem of Berbee (Theorem A) and on the following author's estimate: \(\beta({\mathfrak A},{\mathfrak B})\leq(8N)^{1/2}\alpha({\mathfrak A},{\mathfrak B}),\) the \(\sigma\)- field \({\mathfrak B}\) being purely atomic with exactly N atoms. It is shown that the last estimate is sharp up to a constant factor even if \(\alpha\) (\({\mathfrak A},{\mathfrak B})\) is replaced by \(\rho\) (\({\mathfrak A},{\mathfrak B})\).