On the approximation of P-P and Q-Q plot processes by Brownian bridges (Q1812871)

In the classical statistical problem of the comparison of two independent samples \(X_ 1,\dots,X_ m\) and \(Y_ 1,\dots,Y_ n\) from the distribution functions \(F(x)=P(X_ 1\leq x)\) and \(G(x)=P(Y_ 1\leq x)\), it is now common practice to use quantile-quantile (Q-Q), procentile- procentile (P-P), as well as procentile-quantile (P-Q) plots for data analysis. The Q-Q plots and P-P plots are defined as follows. Let \[ F^ \leftarrow(s)=\inf\{x: F(x)\geq s\}\quad\text{and}\quad G^ \leftarrow(s)=\inf\{x: G(x)\geq s\} \] denote for \(0<s<1\) the (left- continuous versions of the) quantile functions of \(F\) and \(G\) (we use the notation \(F^ \leftarrow\) and \(G^ \leftarrow\) rather than \(F^{-1}\) and \(G^{-1}\) to avoid any ambiguity). Set \[ A=\inf\{x: G(x)>0\}\quad\text{and}\quad B=\sup\{x: G(x)<1\}. \] The Q-Q plot of \(F\) against \(G\) is given by \[ \delta_{FG}(x)=F^ \leftarrow(G(x))\quad\text{for}\quad A<x<B. \] The P-P plot of \(F\) against \(G\) is likewise \[ \tilde\delta_{FG}(s)=F(G^ \leftarrow(s))\quad\text{for}\quad 0<s<1. \] We establish optimal approximation rates of empirical P-P and Q-Q plot processes by sequences of Brownian bridges.