Approximating the inverse of a symmetric matrix with non-negative elements

For an $n\times n$ symmetric positive definite matrix $T=(t_{i,j})$ with positive elements satisfying $t_{i,i}\ge \sum_{j\neq i} t_{i,j}$ and certain bounding conditions, we propose to use the matrix $S=(s_{i,j})$ to approximate its inverse, where $s_{i,j}=\delta_{i,j}/t_{i,i}-1/t_{..}$, $\delta_{i,j}$ is the Kronecker delta function, and $t_{..}=\sum_{i,j=1}^{n}(1-\delta_{i,j}) t_{i,j}$. An explicit bound on the approximation error is obtained, showing that the inverse is well approximated to order $1/(n-1)^2$ uniformly. The results are further extended to allow some off-diagonal elements of $T$ to be zeros.