Distance is determined by taking the results of each calculation and subtracting
them from 1. Thus, the largest distance value for these two subjects is
associated with the Russell/Rao index, 1−2/7 = 3/7, while the smallest distance
is associated with the Matching and Dice coefficients, 1 − 4/7 = 3/7. After distances
are calculated for an entire set of data, they are combined into a matrix
that is entered into a standard clustering algorithm such as Ward’s (Ward, 1963).
3. Methodology
In order to compare the performance of these indices in terms of correctly
grouping individuals, a set of Monte Carlo simulations were conducted under
a variety of conditions, and the 4 distance measures, along with the raw data
method, were applied to assess their performance. The data for this Monte Carlo
study were generated using a 2-parameter logistic (2PL) model, which takes the
following form: