Statistical modeling implications for coffee progenies selection