Ecologists commonly quantify abundances of multiple interacting species and associated abiotic variables in study sites over time. Monitoring programs, shorter-term field studies, and experimental studies all may generate such multispecies and multivariate time series. In spite of the prevalence of this type of data collection, most ecologists have not been trained in time-series analyses appropriate to these community time-series data. Temporal autocorrelation violates the independence assumptions of common statistical tools, and direct and indirect effects complicate interpretations. To cope, many researchers aggregate their data to a level that minimizes such “problems” but also discards a great deal of systemic detail that could be useful to researchers and managers. Multispecies (or Multivariate) Autoregressive (MAR) modeling is a versatile modeling approach that allows users to analyze time-series data and address many common questions in ecology, such as: 1) Which abiotic and biotic factors best predict the abundance of various taxa?; 2) What are the relative strengths of abiotic and biotic interactions between community members?; and 3) How stable is the community structure? In this workshop, we will give participants a basic introduction to autoregressive time-series modeling in general and to multivariate autoregressive modeling in particular. We will cover the steps of model selection, point estimation, and uncertainty specification required for this maximum-likelihood approach. To complement the technical discussions, case studies will be presented to show how MARs have been used to analyze species interactions, environmental drivers, and community stability in aquatic ecosystems. In the afternoon, participants will be introduced to LAMBDA, a free, open-source MAR analysis toolbox, and we will work through the analysis of a variety of datasets. By the end of the workshop, participants should have a working knowledge of multivariate autoregressive models, understand the steps required during application of these methods, and be able to do a basic analysis using LAMBDA. No programming knowledge is required. A working knowledge of multiple regression will be helpful but not necessary. Participants should bring laptops, as no computers will be available. Participants without laptops will be partnered with a laptop-enabled participant for the computer labs in the afternoon.