Given a dataset $X$ with $n$ elements, arithmetic mean $A$, and geometric mean $G$, what is the relationship between the variance of the logs $\sigma_{\ln X}^2$ and these means when the values are very close to $A$?