On Identifying Total Effects in the Presence of Latent Variables and Selection bias