In fact, from the 55 studied parameters,
only the listed 8 contributed significantly to separate the assayed
T levels. Furthermore, the classification performance was not
100% accurate, since 15.2% of the ‘‘5 C” samples were classified
as ‘‘25 C” and 19.4% the ‘‘25 C” samples were classified as ‘‘5 C”
for the original grouped cases, while 21.2% of the ‘‘5 C” samples
were classified as ‘‘25 C” and 22.2% the ‘‘25 C” samples were classified
as ‘‘5 C” for the cross-validated grouped cases.