Religious attendance after elevated depressive symptoms: is selection bias at work?