A law of the iterated logarithm for double array sum of \(\varphi\)-mixing sequence (Q1179953)

Let \(\{\varepsilon_ i\}\) be a bounded \(\varphi\)-mixing sequence and \(\{a_{ni}\}\) a double array of constants. By using Bernstein ad Utev's inequality the author shows \[ \limsup_ n|\sum^ n_{i=1}a_{ni}\varepsilon_ i|/(A_ n\log\log A_ n)^{1/2}\leq b<\infty \hbox{ a.s.}, \] where \(A_ n=\sum^ n_{i=1}a^ 2_{ni}\), and \(b\) is some constant, under certain conditions on \(\{\varepsilon_ i\}\) and \(\{a_{ni}\}\). From this result one can get the most powerful convergence rate of \(LS\) estimates in linear models.