On estimation and influence diagnostics for zero-inflated negative binomial regression models