PS 490: Time Series and Panel Methods

Social scientists routinely attempt statistical and causal inferences using data from a single case in which key variables of interest change over time (time series models), and also using data from multiple cases over time (pooled time series or panel models). These can be powerful techniques; they can sometimes provide more evidence than purely cross-sectional designs, and they can rule out some common inferential problems. However, examining data over time creates its own set of statistical and substantive challenges, and a rich methodological literature has developed around solutions.

In this course we will look at the problems that have been discovered with, and the solutions that have been proposed for, both time series and pooled time series. The first part of the class will be spent on single-case time series, both because they are important in themselves and because the issues that arise within a single case carry over to pooled time series frameworks. We will discuss common time-series data structures, review statistical diagnostics, and cover a number of models that can be used to model time series data, such as autoregressive distributed lag, error-correction, and vector autoregressive models. In the second part of the course, we will explore the additional issues that arise when pooling time series data across multiple units. Much of the discussion during this part involves issues related to temporal and spatial heterogeneity.

In the second part of the course, we will explore the additional issues that arise when pooling time series data across multiple units. Much of the discussion during this part involves issues related to temporal and spatial heterogeneity.

By the end of this course you should be able to: 1) Describe the rationale behind, and the statistical structure of, a variety of time series and pooled time series models. 2) Test for common issues specific to time-series and panel data. 3) Be able to offer informed critiques of articles and books utilizing time-series and panel methods. 4) Produce original research using time-series and panel data and models.