Dependencies among observations are of interest to psychologists for b
oth methodological and substantive reasons. A general procedure is dev
eloped that can be used to estimate dependencies among observations, g
iven a set of a priori expectations about the structure of those depen
dencies. The utility of this general procedure is illustrated for spec
ific cases: the estimation of serial dependency in time-series data, o
f dyadic and group interdependence in social psychology, and of reliab
ility. The utility of the approach derives from its generality and fro
m its ability to handle randomly missing data.