The Beta Version: On why we use "log" likelihood estimations