The U.S. Pacific groundfish community has been studied by NOAA since 1977 in standardized surveys for stock assessment purposes . This long-term dataset consists of an intensive triennial survey from the U.S.-Canada border to southern California, at a variety of depths from 50 – 400 m. Using multivariate auto-regressive first-order (MAR-1) models, with upwelling index as our physical covariate, we estimated the interaction strengths among the components of the groundfish community. We used catch-per-unit-effort (CPUE) data based on both biomass and numerical counts for the 38 most abundant species throughout the region, and examined population dynamics for the larger taxonomic groups (e.g., Family, Order). We also used the maximum eigenvalue of the interaction matrix to estimate the stability of the system. In this presentation, we will compare these estimates by depth and latitude, and discuss the differences in conclusions drawn from using biomass vs. numerical count data. We will examine which populations are most strongly affected by the physical covariate, and which are most sensitive to perturbations. We conclude by describing the lessons learned from performing this analysis, and the strengths and weaknesses of using the MAR-1 approach on datasets such as the ones we present.