Our problem diers from those addressed by time series research or traditional statistical methods
such as cohort or cross-sectional analyses (Glenn, 1977; Menard, 1991). In time series work, we
typically have observations spaced through time with one observation per time point. In contrast, we
have multiple observations at a few discrete points in time. For example, we could have thousands
of observations for each of the 1970, 1980, and 1990 federal censuses. Cohort and cross-sectional
analyses are typically guided by well dened prior hypotheses determined by domain knowledge. For
example, a research question might be How does aging aect political party aliation?" In contrast, our work would take descriptions of young and old people and return all dierences between
them, one of which could be a dierence in political aliation