Because the pattern for the Kappa coefficient and the percent correctly grouped
is virtually identical for the five methods used in this paper, the results of Kappa
are emphasized in subsequent discussion, though the percent correctly grouped
is also presented. This choice is made based on the fact that researchers have
identified relevant cut off values for Kappa (i.e., what values represent high agreement
and what values represent low agreement), making it easier to interpret
and generalize than the percent correctly classified. The effect of differences in