Estimating the causal effect of gun prevalence on homicide rates: a local average treatment effect approach (Discussion Paper)