1: \begin{abstract} In this article, we provide an extension of the Chen--Stein inequality for Poisson approximation in the total variation distance for sums of independent Bernoulli random variables in two ways. We prove that:

2: \begin{itemize}

3: \item we can improve the rate of convergence (hence the quality of the approximation) by using explicitly constructed signed or positive probability measures;

4: \item we can extend the setting to possibly dependent random variables.

5: \end{itemize}

6: The framework which allows this is that of mod-Poisson convergence and more precisely those mod-Poisson convergent sequences whose residue functions can be expressed as a specialization of the generating series of elementary symmetric functions. This combinatorial reformulation allows us to have a general and unified framework in which we can fit the classical setting of sums of independent Bernoulli random variables as well as other examples coming e.g. from probabilistic number theory and random permutations.

7: \end{abstract}

8: