1: \begin{definition}[G-normal distribution]\label{def:gnormal}

2: Let $\cP_Z$ be a set of  probability measures defined on the space $(\Omega, \cF)$.

3: A measurable function $Z\colon  \Omega \mapsto \bbR$  is said to follow a G-normal distribution with lower variance $\smin^2$ and upper variance $\smax^2$ $(0 \leq \smin \leq \smax)$,   if, for every Lipschitz function $\varphi$, 

4: \begin{equation*}\label{eq:def.G-normal}

5: %\cE [ \varphi(Z)] \define  

6: \sup\limits_{\pr  \in \cP_Z} \E_{\pr} [ \varphi(Z) ] =  \sup\limits_{\pr  \in \cP_Z} \int_\Omega \varphi(Z) \, d \pr  = u(1, 0; \varphi),

7: \end{equation*}

8: where $u(1, 0; \varphi)$ is as given in Theorem~\ref{th:clt}. 

9: \end{definition}

10: