Does ignoring model selection when assessing the effect of particulate matter air pollution on mortality make us too vigilant?