Validation of sociometric structure by data on individuals' attributes

Longstanding challenges to relate sociometric structure of an intact social group to other social variables are answered first by obtaining a partition of a set of 69 inmates in Cook County Jail, using a clustering algorithm (CONCOR) applied to sociometric data. We then demonstrate that the resulting partition can be accounted for by substantively interpretable variables (e.g. number of years of school prior to incarceration) measured on the individual inmates. The linear discriminant function used to relate the two types of data does a good job of accounting for processes generating sociometric ties of friendship. It is argued that this methodology can be used to study the diversity of inmate organizational structure.